Daily Papers

The inference cost of Large Language Models (LLMs) has become a critical factor in determining their commercial viability and widespread adoption. This paper introduces a quantitative ``economics of inference'' framework, treating the LLM inference process as a compute-driven intelligent production activity. We analyze its marginal cost, economies of scale, and quality of output under various performance configurations. Based on empirical data from WiNEval-3.0, we construct the first ``LLM Inference Production Frontier,'' revealing three principles: diminishing marginal cost, diminishing returns to scale, and an optimal cost-effectiveness zone. This paper not only provides an economic basis for model deployment decisions but also lays an empirical foundation for the future market-based pricing and optimization of AI inference resources.

The deployment of Large Language Models (LLMs) on edge devices is fundamentally constrained by the "Memory Wall" the bottleneck where data movement latency outstrips arithmetic throughput. Standard inference runtimes often incur significant overhead through high-level abstractions, dynamic dispatch, and unaligned memory access patterns. In this work, we present a novel "Virtual Tensor Core" architecture implemented in software, optimized specifically for ARM64 microarchitectures (Apple Silicon). By bypassing standard library containers in favor of direct memory mapping (mmap) and implementing hand-tuned NEON SIMD kernels, we achieve a form of "Software-Defined Direct Memory Access (DMA)." Our proposed Tensor Virtualization Layout (TVL) guarantees 100% cache line utilization for weight matrices, while our zero-copy loader eliminates initialization latency. Experimental results on a 110M parameter model demonstrate a stable throughput of >60 tokens/second on M2 hardware. While proprietary hardware accelerators (e.g., Apple AMX) can achieve higher peak throughput, our architecture provides a fully open, portable, and deterministic reference implementation for studying the memory bottleneck on general-purpose ARM silicon, meeting the 200ms psycholinguistic latency threshold without opaque dependencies.

Physical AI systems, including robots, autonomous vehicles, embodied agents and edge copilots, often run a different inference workload from cloud LLM serving: single-stream, batch-1 autoregressive decode, where one robot, camera feed or user session waits on the next token. This workload is usually described as memory-bandwidth-bound. Each decode step streams model weights and the active KV cache, so latency should scale with peak HBM bandwidth. We show that this account is true but incomplete. We measure batch-1 decode for three 7 to 8B-class GQA transformers across four NVIDIA GPUs: H100 SXM5, A100-80GB SXM4, L40S and L4. We evaluate context lengths from 2048 to 16384, producing 44 valid cells under a controlled bf16 SDPA setup. The achieved fraction of peak HBM bandwidth falls as peak bandwidth rises. On the headline Qwen-2.5-7B ctx=2048 cell, an L4 reaches roughly 81 percent of its analytic memory floor, while an H100 reaches only 27 percent. Physical-AI decode is memory-dominated, but faster memory does not translate into proportional latency gains. We test the missing term with a CUDA Graphs A/B experiment. On H100 at ctx=2048, CUDA Graphs improves decode latency by 1.259x across N=10 fresh sessions, with a 95 percent bootstrap confidence interval of 1.253 to 1.267. On L4, the same intervention gives only 1.028x. This isolates a launch-side overhead that becomes visible on fast GPUs but remains mostly hidden on slower, bandwidth-bound GPUs. The deployment implication is that memory savings matter only when the runtime realises them. On L4, bf16 decode sits close to the memory floor, but common quantised paths do not recover the expected 4x weight-traffic reduction: bnb-nf4 reaches 59.36 ms/step and AutoAWQ+Marlin reaches 45.24 ms/step from a 62.32 ms bf16 baseline. GPTQ+ExLlamaV2, with Ada-tuned int4 kernels, reaches 17.36 ms/step.

Multi-channel sensor networks in industrial IoT often exceed available bandwidth. We propose PCA-Triage, a streaming algorithm that converts incremental PCA loadings into proportional per-channel sampling rates under a bandwidth budget. PCA-Triage runs in O(wdk) time with zero trainable parameters (0.67 ms per decision). We evaluate on 7 benchmarks (8--82 channels) against 9 baselines. PCA-Triage is the best unsupervised method on 3 of 6 datasets at 50% bandwidth, winning 5 of 6 against every baseline with large effect sizes (r = 0.71--0.91). On TEP, it achieves F1 = 0.961 +/- 0.001 -- within 0.1% of full-data performance -- while maintaining F1 > 0.90 at 30% budget. Targeted extensions push F1 to 0.970. The algorithm is robust to packet loss and sensor noise (3.7--4.8% degradation under combined worst-case).

The rapid expansion of distributed Artificial Intelligence (AI) workloads beyond centralized data centers creates a demand for new communication substrates. These substrates must operate reliably in heterogeneous and permissionless environments, where Network Address Translators (NATs) and firewalls impose significant constraints. Existing solutions, however, are either designed for controlled data center deployments or implemented as monolithic systems that tightly couple machine learning logic with networking code. To address these limitations, we present Lattica, a decentralized cross-NAT communication framework designed to support distributed AI systems. Lattica integrates three core components. First, it employs a robust suite of NAT traversal mechanisms to establish a globally addressable peer-to-peer mesh. Second, it provides a decentralized data store based on Conflict-free Replicated Data Types (CRDTs), ensuring verifiable and eventually consistent state replication. Third, it incorporates a content discovery layer that leverages distributed hash tables (DHTs) together with an optimized RPC protocol for efficient model synchronization. By integrating these components, Lattica delivers a complete protocol stack for sovereign, resilient, and scalable AI systems that operate independently of centralized intermediaries. It is directly applicable to edge intelligence, collaborative reinforcement learning, and other large-scale distributed machine learning scenarios.

AI agent inference is driving an inference heavy datacenter future and exposes bottlenecks beyond compute - especially memory capacity, memory bandwidth and high-speed interconnect. We introduce two metrics - Operational Intensity (OI) and Capacity Footprint (CF) - that jointly explain regimes the classic roofline analysis misses, including the memory capacity wall. Across agentic workflows (chat, coding, web use, computer use) and base model choices (GQA/MLA, MoE, quantization), OI/CF can shift dramatically, with long context KV cache making decode highly memory bound. These observations motivate disaggregated serving and system level heterogeneity: specialized prefill and decode accelerators, broader scale up networking, and decoupled compute-memory enabled by optical I/O. We further hypothesize agent-hardware co design, multiple inference accelerators within one system, and high bandwidth, large capacity memory disaggregation as foundations for adaptation to evolving OI/CF. Together, these directions chart a path to sustain efficiency and capability for large scale agentic AI inference.

AI agents that execute tasks via tool calls frequently hallucinate results - fabricating tool executions, misstating output counts, or presenting inferences as facts. Recent approaches to verifiable AI inference rely on zero-knowledge proofs, which provide cryptographic guarantees but impose minutes of proving time per query, making them impractical for interactive agents. We propose NabaOS, a lightweight verification framework inspired by Indian epistemology (Nyaya Shastra), which classifies every claim in an LLM response by its epistemic source (pramana): direct tool output (pratyaksha), inference (anumana), external testimony (shabda), absence (abhava), or ungrounded opinion. Our runtime generates HMAC-signed tool execution receipts that the LLM cannot forge, then cross-references claims against these receipts to detect hallucinations in real time. We evaluate on NyayaVerifyBench, a new benchmark of 1,800 agent response scenarios across four languages with injected hallucinations of six types. NabaOS detects 94.2% of fabricated tool references, 87.6% of count misstatements, and 91.3% of false absence claims, with <15ms verification overhead per response. For deep delegation (agents performing multi-step web tasks), our cross-checking protocol catches 78.4% of URL fabrications via independent re-fetching. We compare against five approaches: zkLLM (cryptographic proofs, 180s/query), TOPLOC (locality-sensitive hashing), SPEX (sampling-based proof of execution), tensor commitments, and self-consistency checking. NabaOS achieves the best cost-latency-coverage trade-off for interactive agents: 94.2% coverage at <15ms versus zkLLM's near-perfect coverage at 180,000ms. For interactive agents, practical receipt-based verification provides better cost-benefit than cryptographic proofs, and epistemic classification gives users actionable trust signals rather than binary judgments.

The integration of machine learning (ML) systems into critical industries such as healthcare, finance, and cybersecurity has transformed decision-making processes, but it also brings new challenges around trust, security, and accountability. As AI systems become more ubiquitous, ensuring the transparency and correctness of AI-driven decisions is crucial, especially when they have direct consequences on privacy, security, or fairness. Verifiable AI, powered by Zero-Knowledge Machine Learning (zkML), offers a robust solution to these challenges. zkML enables the verification of AI model inferences without exposing sensitive data, providing an essential layer of trust and privacy. However, traditional zkML systems typically require deep cryptographic expertise, placing them beyond the reach of most ML engineers. In this paper, we introduce JSTprove, a specialized zkML toolkit, built on Polyhedra Network's Expander backend, to enable AI developers and ML engineers to generate and verify proofs of AI inference. JSTprove provides an end-to-end verifiable AI inference pipeline that hides cryptographic complexity behind a simple command-line interface while exposing auditable artifacts for reproducibility. We present the design, innovations, and real-world use cases of JSTprove as well as our blueprints and tooling to encourage community review and extension. JSTprove therefore serves both as a usable zkML product for current engineering needs and as a reproducible foundation for future research and production deployments of verifiable AI.

In this paper, we propose a novel approach where agents can form swarms to produce high-quality responses effectively. This is accomplished by utilizing agents capable of data inference and ranking, which can be effectively implemented using LLMs as response classifiers. We assess existing approaches for trustless agent inference, define our methodology, estimate practical parameters, and model various types of malicious agent attacks. Our method leverages the collective intelligence of swarms, ensuring robust and efficient decentralized AI inference with better accuracy, security, and reliability. We show that our approach is an order of magnitude faster than other trustless inference strategies reaching less than 125 ms validation latency.

The rapid growth of large-scale AI models, particularly large language models has brought significant challenges in data privacy, computational resources, and accessibility. Traditional centralized architectures often struggle to meet required data security and scalability needs which hinders the democratization of AI systems. Nesa introduces a model-agnostic sharding framework designed for decentralized AI inference. Our framework uses blockchain-based sequential deep neural network sharding to distribute computational tasks across a diverse network of nodes based on a personalised heuristic and routing mechanism. This enables efficient distributed training and inference for recent large-scale models even on consumer-grade hardware. We use compression techniques like dynamic blockwise quantization and mixed matrix decomposition to reduce data transfer and memory needs. We also integrate robust security measures, including hardware-based trusted execution environments to ensure data integrity and confidentiality. Evaluating our system across various natural language processing and vision tasks shows that these compression strategies do not compromise model accuracy. Our results highlight the potential to democratize access to cutting-edge AI technologies by enabling secure and efficient inference on a decentralized network.

As centralized AI hits compute ceilings and diminishing returns from ever-larger training runs, meeting demand requires an inference layer that scales horizontally in both capacity and capability. We present Fortytwo, a novel protocol that leverages swarm intelligence principles and distributed pairwise ranking consensus to achieve superior performance in AI inference. Our approach reimagines collaboration among AI nodes using swarm inference: a peer-ranked, reputation-weighted consensus across heterogeneous models that surfaces the highest-quality responses. Using pairwise ranking with a custom Bradley-Terry-style aggregation model, we demonstrate that swarm inference substantially outperforms majority voting, achieving 85.90% on GPQA Diamond versus 68.69% for majority voting with the same model set - an improvement of +17.21 percentage points (approximately +25.1% relative). The protocol incorporates on-chain reputation so node influence adapts to demonstrated accuracy over time, yielding a meritocratic consensus that filters low-quality or malicious participants. To resist Sybil attacks, Fortytwo employs proof-of-capability in its consensus: nodes must successfully complete calibration/test requests and stake reputation to enter ranking rounds, making multi-identity attacks economically unattractive while preserving openness. Across six challenging benchmarks, including GPQA Diamond, LiveCodeBench, and AIME, our evaluation indicates higher accuracy and strong resilience to adversarial and noisy free-form prompting (e.g., prompt-injection degradation of only 0.12% versus 6.20% for a monolithic single-model baseline), while retaining practical deployability. Together, these results establish a foundation for decentralized AI systems - democratizing access to high-quality inference through collective intelligence without sacrificing reliability or security.

The rapid adaptation of data driven AI models, such as deep learning inference, training, Vision Transformers (ViTs), and other HPC applications, drives a strong need for runtime precision configurable different non linear activation functions (AF) hardware support. Existing solutions support diverse precision or runtime AF reconfigurability but fail to address both simultaneously. This work proposes a flexible and SIMD multiprecision processing element (FlexPE), which supports diverse runtime configurable AFs, including sigmoid, tanh, ReLU and softmax, and MAC operation. The proposed design achieves an improved throughput of up to 16X FxP4, 8X FxP8, 4X FxP16 and 1X FxP32 in pipeline mode with 100% time multiplexed hardware. This work proposes an area efficient multiprecision iterative mode in the SIMD systolic arrays for edge AI use cases. The design delivers superior performance with up to 62X and 371X reductions in DMA reads for input feature maps and weight filters in VGG16, with an energy efficiency of 8.42 GOPS / W within the accuracy loss of 2%. The proposed architecture supports emerging 4-bit computations for DL inference while enhancing throughput in FxP8/16 modes for transformers and other HPC applications. The proposed approach enables future energy-efficient AI accelerators in edge and cloud environments.

This document presents a vision for a novel AI infrastructure design that has been initially validated through inference simulations on state-of-the-art large language models. Advancements in deep learning and specialized hardware have driven the rapid growth of large language models (LLMs) and generative AI systems. However, traditional GPU-centric architectures face scalability challenges for inference workloads due to limitations in memory capacity, bandwidth, and interconnect scaling. To address these issues, the FengHuang Platform, a disaggregated AI infrastructure platform, is proposed to overcome memory and communication scaling limits for AI inference. FengHuang features a multi-tier shared-memory architecture combining high-speed local memory with centralized disaggregated remote memory, enhanced by active tensor paging and near-memory compute for tensor operations. Simulations demonstrate that FengHuang achieves up to 93% local memory capacity reduction, 50% GPU compute savings, and 16x to 70x faster inter-GPU communication compared to conventional GPU scaling. Across workloads such as GPT-3, Grok-1, and QWEN3-235B, FengHuang enables up to 50% GPU reductions while maintaining end-user performance, offering a scalable, flexible, and cost-effective solution for AI inference infrastructure. FengHuang provides an optimal balance as a rack-level AI infrastructure scale-up solution. Its open, heterogeneous design eliminates vendor lock-in and enhances supply chain flexibility, enabling significant infrastructure and power cost reductions.

Modern cutting-edge AI applications are being developed over fast-evolving, heterogeneous, nascent hardware devices. This requires frequent reworking of the AI software stack to adopt bottom-up changes from new hardware, which takes time for general-purpose software libraries. Consequently, real applications often develop custom software stacks optimized for their specific workloads and hardware. Custom stacks help in quick development and optimization, but incur a lot of redundant efforts across applications in writing non-portable code. This paper discusses an alternative communication library interface for AI applications that offers both portability and performance by reducing redundant efforts while maintaining flexibility for customization. We present MSCCL++, a novel abstraction of GPU communication based on separation of concerns: (1) a primitive interface provides a minimal hardware abstraction as a common ground for software and hardware developers to write custom communication, and (2) higher-level portable interfaces and specialized implementations enable optimization for different workloads and hardware environments. This approach makes the primitive interface reusable across applications while enabling highly flexible optimization. Compared to state-of-the-art baselines (NCCL, RCCL, and MSCCL), MSCCL++ achieves speedups of up to 5.4times for collective communication and up to 15% for real-world AI inference workloads. MSCCL++ is in production of multiple AI services provided by Microsoft Azure, and is also adopted by RCCL, the GPU collective communication library maintained by AMD. MSCCL++ is open-source and available at https://github.com/microsoft/mscclpp.

The design space of agentic AI inference spans two extremes: frontier large language models (LLMs), typically hosted in the cloud and offering strong performance across a wide range of tasks at substantially high cost, and more cost-efficient small language models (SLMs), which are amenable to on-device inference. Hybrid multi-agent systems (MASs) combining on-device and cloud models offer a promising middle ground, but they also introduce a complex and poorly understood design space in which task accuracy, monetary cost, and edge energy consumption are tightly coupled; in the absence of general design principles, hybrid components, although not the most prevalent choice, are typically introduced through ad hoc decisions tailored to specific domains. In this work, we examine this design space more systematically. We adapt two representative MAS architectures to support hybrid inference and study how individual design choices shift the operating point along the Pareto frontier of power, cost, and performance. Our findings paint a nuanced picture of hybrid MAS design: while SLMs can effectively benefit from LLM assistance, the optimal architecture is highly task-dependent, and greater frontier-level compute does not consistently translate to better performance.

The AIPC concept is gaining popularity, and more and more hybrid CPUs will be running AI models on client devices. However, the current AI inference framework overlooks the imbalanced hardware capability of hybrid CPUs, leading to low inference performance. To address this issue, we have introduced a dynamic parallel method for hybrid CPUs, which significantly increases LLM inference performance by balancing the workload for each core of a hybrid CPU before the parallel work starts. This method has enabled Neural Speed to achieve more than 90% (on average) of memory bandwidth on two hybrid Intel CPUs.

Narrow bit-width data formats are key to reducing the computational and storage costs of modern deep learning applications. This paper evaluates Microscaling (MX) data formats that combine a per-block scaling factor with narrow floating-point and integer types for individual elements.MX formats balance the competing needs of hardware efficiency, model accuracy, and user friction. Empirical results on over two dozen benchmarks demonstrate practicality of MX data formats as a drop-in replacement for baseline FP32 for AI inference and training with low user friction. We also show the first instance of training generative language models at sub-8-bit weights, activations, and gradients with minimal accuracy loss and no modifications to the training recipe.

Automated ad targeting on social media is opaque, creating risks of exploitation and invisibility to external scrutiny. Users may be steered toward harmful content while independent auditing of these processes remains blocked. Large Language Models (LLMs) raise a new concern: the potential to reverse-engineer sensitive user attributes from exposure alone. We introduce a multi-stage auditing framework to investigate these risks. First, a large-scale audit of over 435,000 ad impressions delivered to 891 Australian Facebook users reveals algorithmic biases, including disproportionate Gambling and Politics ads shown to socioeconomically vulnerable and politically aligned groups. Second, a multimodal LLM can reconstruct users' demographic profiles from ad streams, outperforming census-based baselines and matching or exceeding human performance. Our results provide the first empirical evidence that ad streams constitute rich digital footprints for public AI inference, highlighting urgent privacy risks and the need for content-level auditing and governance.

AI agents execute tens to hundreds of chained LLM calls per task, yet GPU schedulers treat each call as independent, discarding gigabytes of intermediate state between steps and inflating end-to-end latency by 3-8x. We argue that this request-level abstraction is fundamentally mismatched to compound AI workloads, and propose a shift to program-level scheduling: treating the entire agent workflow (not individual inference calls) as the first-class schedulable unit. We present SAGA, a distributed scheduler that implements this abstraction through three mechanisms: (1) Agent Execution Graphs that capture workflow structure to predict KV cache reuse across tool-call boundaries, achieving within 1.31x of Bélády's optimal offline policy; (2) session-affinity batching with work stealing that co-locates correlated requests while maintaining global load balance; and (3) Agent Fair Share, a task-completion-time fairness metric with provable bounded-deviation guarantees. On a 64-GPU cluster serving SWE-bench coding agents and WebArena browser tasks, SAGA reduces task completion time by 1.64x (geometric mean, p < 0.001) over vLLM v0.15.1 with prefix caching and affinity routing, while improving GPU memory utilization by 1.22x and achieving 99.2% SLO attainment under multi-tenant interference. These latency gains come at a quantified cost: approximately 30% lower peak throughput than throughput-optimal batch scheduling, a tradeoff appropriate for the latency-sensitive interactive deployments that dominate compound AI usage. Our results demonstrate that workflow-aware scheduling is essential for efficient compound AI serving.

Recent advances in artificial intelligence (AI), particularly deep learning, have led to widespread adoption across various applications. Yet, a fundamental challenge persists: how can we verify the correctness of AI model inference when model owners cannot (or will not) reveal their parameters? These parameters represent enormous training costs and valuable intellectual property, making transparent verification difficult. In this paper, we introduce a zero-knowledge framework capable of verifying deep learning inference without exposing model internal parameters. Built on recursively composed zero-knowledge proofs and requiring no trusted setup, our framework supports both linear and nonlinear neural network layers, including matrix multiplication, normalization, softmax, and SiLU. Leveraging the Fiat-Shamir heuristic, we obtain a succinct non-interactive argument of knowledge (zkSNARK) with constant-size proofs. To demonstrate the practicality of our approach, we translate the DeepSeek model into a fully SNARK-verifiable version named ZK-DeepSeek and show experimentally that our framework delivers both efficiency and flexibility in real-world AI verification workloads.

AI computing and data centers consume a large amount of freshwater, both directly for cooling and indirectly for electricity generation. While most attention has been paid to developed countries such as the U.S., this paper presents the first-of-its-kind dataset that combines nation-level weather and electricity generation data to estimate water usage efficiency for data centers in 41 African countries across five different climate regions. We also use our dataset to evaluate and estimate the water consumption of inference on two large language models (i.e., Llama-3-70B and GPT-4) in 11 selected African countries. Our findings show that writing a 10-page report using Llama-3-70B could consume about 0.7 liters of water, while the water consumption by GPT-4 for the same task may go up to about 60 liters. For writing a medium-length email of 120-200 words, Llama-3-70B and GPT-4 could consume about 0.13 liters and 3 liters of water, respectively. Interestingly, given the same AI model, 8 out of the 11 selected African countries consume less water than the global average, mainly because of lower water intensities for electricity generation. However, water consumption can be substantially higher in some African countries with a steppe climate than the U.S. and global averages, prompting more attention when deploying AI computing in these countries. Our dataset is publicly available on https://huggingface.co/datasets/masterlion/WaterEfficientDatasetForAfricanCountries/tree/main{Hugging Face}.

Researchers have long recognized the socio-technical gaps in personal tracking research, where machines can never fully model the complexity of human behavior, making it only able to produce basic rule-based outputs or "black-box" results that lack clear explanations. Real-world deployments rely on experts for this complex translation from sparse data to meaningful insights. In this study, we consider this translation process from data to insights by experts as "sensemaking" and explore how HCI researchers can support it through Vital Insight, an evidence-based 'sensemaking' system that combines direct representation and indirect inference through visualization and Large Language Models. We evaluate Vital Insight in user testing sessions with 14 experts in multi-modal tracking, synthesize design implications, and develop an expert sensemaking model where they iteratively move between direct data representations and AI-supported inferences to explore, retrieve, question, and validate insights.

Powerful foundation models, including large language models (LLMs), with Transformer architectures have ushered in a new era of Generative AI across various industries. Industry and research community have witnessed a large number of new applications, based on those foundation models. Such applications include question and answer, customer services, image and video generation, and code completions, among others. However, as the number of model parameters reaches to hundreds of billions, their deployment incurs prohibitive inference costs and high latency in real-world scenarios. As a result, the demand for cost-effective and fast inference using AI accelerators is ever more higher. To this end, our tutorial offers a comprehensive discussion on complementary inference optimization techniques using AI accelerators. Beginning with an overview of basic Transformer architectures and deep learning system frameworks, we deep dive into system optimization techniques for fast and memory-efficient attention computations and discuss how they can be implemented efficiently on AI accelerators. Next, we describe architectural elements that are key for fast transformer inference. Finally, we examine various model compression and fast decoding strategies in the same context.

The deployment of large-scale models, such as large language models (LLMs) and sophisticated image generation systems, incurs substantial costs due to their computational demands. To mitigate these costs and address challenges related to scalability and data security, there is a growing shift towards decentralized systems for deploying such models. In these decentralized environments, efficient inference acceleration becomes crucial to manage computational resources effectively and enhance system responsiveness. In this work, we address the challenge of selecting optimal acceleration methods in decentralized systems by introducing a meta-learning-based framework. This framework automates the selection process by learning from historical performance data of various acceleration techniques across different tasks. Unlike traditional methods that rely on random selection or expert intuition, our approach systematically identifies the best acceleration strategies based on the specific characteristics of each task. We demonstrate that our meta-learning framework not only streamlines the decision-making process but also consistently outperforms conventional methods in terms of efficiency and performance. Our results highlight the potential of meta-learning to revolutionize inference acceleration in decentralized AI systems, offering a path towards more democratic and economically feasible artificial intelligence solutions.

Understanding emotions is fundamental to human interaction and experience. Humans easily infer emotions from situations or facial expressions, situations from emotions, and do a variety of other affective cognition. How adept is modern AI at these inferences? We introduce an evaluation framework for testing affective cognition in foundation models. Starting from psychological theory, we generate 1,280 diverse scenarios exploring relationships between appraisals, emotions, expressions, and outcomes. We evaluate the abilities of foundation models (GPT-4, Claude-3, Gemini-1.5-Pro) and humans (N = 567) across carefully selected conditions. Our results show foundation models tend to agree with human intuitions, matching or exceeding interparticipant agreement. In some conditions, models are ``superhuman'' -- they better predict modal human judgements than the average human. All models benefit from chain-of-thought reasoning. This suggests foundation models have acquired a human-like understanding of emotions and their influence on beliefs and behavior.

Speculative Decoding has gained popularity as an effective technique for accelerating the auto-regressive inference process of Large Language Models (LLMs). However, Speculative Decoding entirely relies on the availability of efficient draft models, which are often lacking for many existing language models due to a stringent constraint of vocabulary incompatibility. In this work we introduce FastDraft, a novel and efficient approach for pre-training and aligning a draft model to any large language model by incorporating efficient pre-training, followed by fine-tuning over synthetic datasets generated by the target model. We demonstrate FastDraft by training two highly parameter efficient drafts for the popular Phi-3-mini and Llama-3.1-8B models. Using FastDraft, we were able to produce a draft with approximately 10 billion tokens on a single server with 8 Intel^circledR Gaudi^circledR 2 accelerators in under 24 hours. Our results show that the draft model achieves impressive results in key metrics of acceptance rate, block efficiency and up to 3x memory bound speed up when evaluated on code completion and up to 2x in summarization, text completion and instruction tasks. We validate our theoretical findings through benchmarking on the latest Intel^circledR Core^{tiny TM} Ultra, achieving a wall-clock time speedup of up to 2x, indicating a significant reduction in runtime. Due to its high quality, FastDraft unlocks large language models inference on AI-PC and other edge-devices.

We revisit the longstanding open problem of implementing Nakamoto's proof-of-work (PoW) consensus based on a real-world computational task T(x) (as opposed to artificial random hashing), in a truly permissionless setting where the miner itself chooses the input x. The challenge in designing such a Proof-of-Useful-Work (PoUW) protocol, is using the native computation of T(x) to produce a PoW certificate with prescribed hardness and with negligible computational overhead over the worst-case complexity of T(cdot) -- This ensures malicious miners cannot ``game the system" by fooling the verifier to accept with higher probability compared to honest miners (while using similar computational resources). Indeed, obtaining a PoUW with O(1)-factor overhead is trivial for any task T, but also useless. Our main result is a PoUW for the task of Matrix Multiplication MatMul(A,B) of arbitrary matrices with 1+o(1) multiplicative overhead compared to naive MatMul (even in the presence of Fast Matrix Multiplication-style algorithms, which are currently impractical). We conjecture that our protocol has optimal security in the sense that a malicious prover cannot obtain any significant advantage over an honest prover. This conjecture is based on reducing hardness of our protocol to the task of solving a batch of low-rank random linear equations which is of independent interest. Since MatMuls are the bottleneck of AI compute as well as countless industry-scale applications, this primitive suggests a concrete design of a new L1 base-layer protocol, which nearly eliminates the energy-waste of Bitcoin mining -- allowing GPU consumers to reduce their AI training and inference costs by ``re-using" it for blockchain consensus, in exchange for block rewards (2-for-1). This blockchain is currently under construction.

How humans and machines make sense of current inputs for relation reasoning and question-answering while putting the perceived information into context of our past memories, has been a challenging conundrum in cognitive science and artificial intelligence. Inspired by human brain's memory system and cognitive architectures, we propose a PMI framework that consists of perception, memory and inference components. Notably, the memory module comprises working and long-term memory, with the latter endowed with a higher-order structure to retain more accumulated knowledge and experiences. Through a differentiable competitive write access, current perceptions update working memory, which is later merged with long-term memory via outer product associations, averting memory overflow and minimizing information conflicts. In the inference module, relevant information is retrieved from two separate memory origins and associatively integrated to attain a more comprehensive and precise interpretation of current perceptions. We exploratively apply our PMI to improve prevailing Transformers and CNN models on question-answering tasks like bAbI-20k and Sort-of-CLEVR datasets, as well as relation calculation and image classification tasks, and in each case, our PMI enhancements consistently outshine their original counterparts significantly. Visualization analyses reveal that memory consolidation, along with the interaction and integration of information from diverse memory sources, substantially contributes to the model effectiveness on inference tasks.

Recent advances in 1-bit Large Language Models (LLMs), such as BitNet and BitNet b1.58, present a promising approach to enhancing the efficiency of LLMs in terms of speed and energy consumption. These developments also enable local LLM deployment across a broad range of devices. In this work, we introduce bitnet.cpp, a tailored software stack designed to unlock the full potential of 1-bit LLMs. Specifically, we develop a set of kernels to support fast and lossless inference of ternary BitNet b1.58 LLMs on CPUs. Extensive experiments demonstrate that bitnet.cpp achieves significant speedups, ranging from 2.37x to 6.17x on x86 CPUs and from 1.37x to 5.07x on ARM CPUs, across various model sizes. The code is available at https://github.com/microsoft/BitNet.

Large Language Models (LLMs) have propelled groundbreaking advancements across several domains and are commonly used for text generation applications. However, the computational demands of these complex models pose significant challenges, requiring efficient hardware acceleration. Benchmarking the performance of LLMs across diverse hardware platforms is crucial to understanding their scalability and throughput characteristics. We introduce LLM-Inference-Bench, a comprehensive benchmarking suite to evaluate the hardware inference performance of LLMs. We thoroughly analyze diverse hardware platforms, including GPUs from Nvidia and AMD and specialized AI accelerators, Intel Habana and SambaNova. Our evaluation includes several LLM inference frameworks and models from LLaMA, Mistral, and Qwen families with 7B and 70B parameters. Our benchmarking results reveal the strengths and limitations of various models, hardware platforms, and inference frameworks. We provide an interactive dashboard to help identify configurations for optimal performance for a given hardware platform.

An introduction to the emerging fusion of machine learning and causal inference. The book presents ideas from classical structural equation models (SEMs) and their modern AI equivalent, directed acyclical graphs (DAGs) and structural causal models (SCMs), and covers Double/Debiased Machine Learning methods to do inference in such models using modern predictive tools.

Mastering commonsense understanding and reasoning is a pivotal skill essential for conducting engaging conversations. While there have been several attempts to create datasets that facilitate commonsense inferences in dialogue contexts, existing datasets tend to lack in-depth details, restate information already present in the conversation, and often fail to capture the multifaceted nature of commonsense reasoning. In response to these limitations, we compile a new synthetic dataset for commonsense reasoning in dialogue contexts using GPT, ConvoSense, that boasts greater contextual novelty, offers a higher volume of inferences per example, and substantially enriches the detail conveyed by the inferences. Our dataset contains over 500,000 inferences across 12,000 dialogues with 10 popular inference types, which empowers the training of generative commonsense models for dialogue that are superior in producing plausible inferences with high novelty when compared to models trained on the previous datasets. To the best of our knowledge, ConvoSense is the first of its kind to provide such a multitude of novel inferences at such a large scale.

With the wide adoption of AI applications, there is a pressing need of enabling real-time neural network (NN) inference on small embedded devices, but deploying NNs and achieving high performance of NN inference on these small devices is challenging due to their extremely weak capabilities. Although NN partitioning and offloading can contribute to such deployment, they are incapable of minimizing the local costs at embedded devices. Instead, we suggest to address this challenge via agile NN offloading, which migrates the required computations in NN offloading from online inference to offline learning. In this paper, we present AgileNN, a new NN offloading technique that achieves real-time NN inference on weak embedded devices by leveraging eXplainable AI techniques, so as to explicitly enforce feature sparsity during the training phase and minimize the online computation and communication costs. Experiment results show that AgileNN's inference latency is >6x lower than the existing schemes, ensuring that sensory data on embedded devices can be timely consumed. It also reduces the local device's resource consumption by >8x, without impairing the inference accuracy.

A recurring challenge of crowdsourcing NLP datasets at scale is that human writers often rely on repetitive patterns when crafting examples, leading to a lack of linguistic diversity. We introduce a novel approach for dataset creation based on worker and AI collaboration, which brings together the generative strength of language models and the evaluative strength of humans. Starting with an existing dataset, MultiNLI for natural language inference (NLI), our approach uses dataset cartography to automatically identify examples that demonstrate challenging reasoning patterns, and instructs GPT-3 to compose new examples with similar patterns. Machine generated examples are then automatically filtered, and finally revised and labeled by human crowdworkers. The resulting dataset, WANLI, consists of 107,885 NLI examples and presents unique empirical strengths over existing NLI datasets. Remarkably, training a model on WANLI improves performance on eight out-of-domain test sets we consider, including by 11% on HANS and 9% on Adversarial NLI, compared to training on the 4x larger MultiNLI. Moreover, it continues to be more effective than MultiNLI augmented with other NLI datasets. Our results demonstrate the promise of leveraging natural language generation techniques and re-imagining the role of humans in the dataset creation process.

The rapid scaling of Large Language Models presents significant challenges for their deployment and inference, particularly on resource-constrained specialized AI hardware accelerators such as Huawei's Ascend NPUs, where weight data transfer has become a critical performance bottleneck. While lossless compression can preserve model accuracy and reduce data volume, existing lossless compression algorithms exhibit extremely low throughput when ported to the Ascend NPU architecture. In this paper, we propose ENEC, a novel lossless compression method specifically customized for AI model weights and optimized for Ascend Neural Processing Units. ENEC adopts a block-based fixed-length encoding scheme and incorporates a series of NPU-specific optimizations: bit-width quantization with hierarchical halving bit-packing, vectorized branch-free integer transformation, and dependency-decoupled intra-segment scan for efficient prefix-sum computation. Experimental results demonstrate that ENEC outperforms existing state-of-the-art NPU compressors in both compression ratio and throughput. Compared to leading GPU solutions, ENEC achieves a 3.43X higher throughput than DietGPU and a 1.12X better compression ratio than nvCOMP. By reducing weight transmission overhead, ENEC significantly improves end-to-end inference performance, achieving up to a 6.3X speedup. On Ascend NPUs, ENEC is the first open-source lossless compression algorithm for model weights that achieves performance comparable to state-of-the-art GPU compressors, offering an effective solution for deploying large-scale AI models.

The rapid advancement of ML models in critical sectors such as healthcare, finance, and security has intensified the need for robust data security, model integrity, and reliable outputs. Large multimodal foundational models, while crucial for complex tasks, present challenges in scalability, reliability, and potential misuse. Decentralized systems offer a solution by distributing workload and mitigating central points of failure, but they introduce risks of unauthorized access to sensitive data across nodes. We address these challenges with a comprehensive framework designed for responsible AI development. Our approach incorporates: 1) Zero-knowledge proofs for secure model verification, enhancing trust without compromising privacy. 2) Consensus-based verification checks to ensure consistent outputs across nodes, mitigating hallucinations and maintaining model integrity. 3) Split Learning techniques that segment models across different nodes, preserving data privacy by preventing full data access at any point. 4) Hardware-based security through trusted execution environments (TEEs) to protect data and computations. This framework aims to enhance security and privacy and improve the reliability and fairness of multimodal AI systems. Promoting efficient resource utilization contributes to more sustainable AI development. Our state-of-the-art proofs and principles demonstrate the framework's effectiveness in responsibly democratizing artificial intelligence, offering a promising approach for building secure and private foundational models.

Modern text-to-image generative models can inadvertently reproduce copyrighted content memorized in their training data, raising serious concerns about potential copyright infringement. We introduce Guardians of Generation, a model agnostic inference time framework for dynamic copyright shielding in AI image generation. Our approach requires no retraining or modification of the generative model weights, instead integrating seamlessly with existing diffusion pipelines. It augments the generation process with an adaptive guidance mechanism comprising three components: a detection module, a prompt rewriting module, and a guidance adjustment module. The detection module monitors user prompts and intermediate generation steps to identify features indicative of copyrighted content before they manifest in the final output. If such content is detected, the prompt rewriting mechanism dynamically transforms the user's prompt by sanitizing or replacing references that could trigger copyrighted material while preserving the prompt's intended semantics. The adaptive guidance module adaptively steers the diffusion process away from flagged content by modulating the model's sampling trajectory. Together, these components form a robust shield that enables a tunable balance between preserving creative fidelity and ensuring copyright compliance. We validate our method on a variety of generative models such as Stable Diffusion, SDXL, and Flux, demonstrating substantial reductions in copyrighted content generation with negligible impact on output fidelity or alignment with user intent. This work provides a practical, plug-and-play safeguard for generative image models, enabling more responsible deployment under real-world copyright constraints. Source code is available at: https://respailab.github.io/gog

As the training of giant dense models hits the boundary on the availability and capability of the hardware resources today, Mixture-of-Experts (MoE) models become one of the most promising model architectures due to their significant training cost reduction compared to a quality-equivalent dense model. Its training cost saving is demonstrated from encoder-decoder models (prior works) to a 5x saving for auto-aggressive language models (this work along with parallel explorations). However, due to the much larger model size and unique architecture, how to provide fast MoE model inference remains challenging and unsolved, limiting its practical usage. To tackle this, we present DeepSpeed-MoE, an end-to-end MoE training and inference solution as part of the DeepSpeed library, including novel MoE architecture designs and model compression techniques that reduce MoE model size by up to 3.7x, and a highly optimized inference system that provides 7.3x better latency and cost compared to existing MoE inference solutions. DeepSpeed-MoE offers an unprecedented scale and efficiency to serve massive MoE models with up to 4.5x faster and 9x cheaper inference compared to quality-equivalent dense models. We hope our innovations and systems help open a promising path to new directions in the large model landscape, a shift from dense to sparse MoE models, where training and deploying higher-quality models with fewer resources becomes more widely possible.

Long-context inference in Large Language Models (LLMs) is bottlenecked by the quadratic computation complexity of attention and the substantial memory footprint of Key-Value (KV) caches. While existing sparse attention mechanisms attempt to mitigate this by exploiting inherent sparsity, they often rely on rigid patterns or aggressive pruning, failing to achieve an optimal balance between efficiency and accuracy. In this paper, we introduce {\bf HyLRA} ({\bf Hy}brid {\bf L}ayer {\bf R}euse {\bf A}ttention), a novel framework driven by layer-wise sparsity profiling. Our empirical analysis uncovers a dual characteristic in attention mechanics: intra-layer sensitivity, where specific layers necessitate full attention to prevent feature distortion, and inter-layer similarity, where consecutive layers share substantial critical tokens. Based on these observations, HyLRA employs an offline dynamic programming approach to derive an optimal layer-wise policy. This hybrid strategy retains full attention for sensitive layers to ensure robustness, while enabling tolerant layers to bypass quadratic calculations by directly reusing top-k indices from preceding layers. This approach allows LLMs to restrict computation to the most critical tokens, effectively overcoming the quadratic bottleneck of dense attention. Extensive evaluations demonstrate that HyLRA improves inference throughput by 6\%--46\% while maintaining comparable performance (with <1% accuracy degradation), consistently outperforming state-of-the-art sparse attention methods. HyLRA is open source at https://anonymous.4open.science/r/unified-cache-management-CF80/{/r/unified-cache-management-CF80/}

We introduce a variational reasoning framework for language models that treats thinking traces as latent variables and optimizes them through variational inference. Starting from the evidence lower bound (ELBO), we extend it to a multi-trace objective for tighter bounds and propose a forward-KL formulation that stabilizes the training of the variational posterior. We further show that rejection sampling finetuning and binary-reward RL, including GRPO, can be interpreted as local forward-KL objectives, where an implicit weighting by model accuracy naturally arises from the derivation and reveals a previously unnoticed bias toward easier questions. We empirically validate our method on the Qwen 2.5 and Qwen 3 model families across a wide range of reasoning tasks. Overall, our work provides a principled probabilistic perspective that unifies variational inference with RL-style methods and yields stable objectives for improving the reasoning ability of language models. Our code is available at https://github.com/sail-sg/variational-reasoning.

Training-time privileged information (PI) can enable language models to succeed on tasks they would otherwise fail, making it a powerful tool for reinforcement learning in hard, long-horizon settings. However, transferring capabilities learned with PI to policies that must act without it at inference time remains a fundamental challenge. We study this problem in the context of distilling frontier models for multi-turn agentic environments, where closed-source systems typically hide their internal reasoning and expose only action trajectories. This breaks standard distillation pipelines, since successful behavior is observable but the reasoning process is not. For this, we introduce π-Distill, a joint teacher-student objective that trains a PI-conditioned teacher and an unconditioned student simultaneously using the same model. Additionally, we also introduce On-Policy Self-Distillation (OPSD), an alternative approach that trains using Reinforcement Learning (RL) with a reverse KL-penalty between the student and the PI-conditioned teacher. We show that both of these algorithms effectively distill frontier agents using action-only PI. Specifically we find that π-Distill and in some cases OPSD, outperform industry standard practices (Supervised finetuning followed by RL) that assume access to full Chain-of-Thought supervision across multiple agentic benchmarks, models, and forms of PI. We complement our results with extensive analysis that characterizes the factors enabling effective learning with PI, focusing primarily on π-Distill and characterizing when OPSD is competitive.

Vision-Language-Action models (VLAs) have demonstrated remarkable performance in robot control. However, they remain fundamentally limited in tasks that require high precision due to their single-inference paradigm. While test-time scaling approaches using external verifiers have shown promise, they require additional training and fail to generalize to unseen conditions. We propose Masking Distribution Guided Selection (MG-Select), a novel test-time scaling framework for VLAs that leverages the model's internal properties without requiring additional training or external modules. Our approach utilizes KL divergence from a reference action token distribution as a confidence metric for selecting the optimal action from multiple candidates. We introduce a reference distribution generated by the same VLA but with randomly masked states and language conditions as inputs, ensuring maximum uncertainty while remaining aligned with the target task distribution. Additionally, we propose a joint training strategy that enables the model to learn both conditional and unconditional distributions by applying dropout to state and language conditions, thereby further improving the quality of the reference distribution. Our experiments demonstrate that MG-Select achieves significant performance improvements, including a 28%/35% improvement in real-world in-distribution/out-of-distribution tasks, along with a 168% relative gain on RoboCasa pick-and-place tasks trained with 30 demonstrations.

Low-rank adaptations (LoRA) are often employed to fine-tune large language models (LLMs) for new tasks. This paper investigates LoRA composability for cross-task generalization and introduces LoraHub, a strategic framework devised for the purposive assembly of LoRA modules trained on diverse given tasks, with the objective of achieving adaptable performance on unseen tasks. With just a few examples from a novel task, LoraHub enables the fluid combination of multiple LoRA modules, eradicating the need for human expertise. Notably, the composition requires neither additional model parameters nor gradients. Our empirical results, derived from the Big-Bench Hard (BBH) benchmark, suggest that LoraHub can effectively mimic the performance of in-context learning in few-shot scenarios, excluding the necessity of in-context examples alongside each inference input. A significant contribution of our research is the fostering of a community for LoRA, where users can share their trained LoRA modules, thereby facilitating their application to new tasks. We anticipate this resource will widen access to and spur advancements in general intelligence as well as LLMs in production. Code will be available at https://github.com/sail-sg/lorahub.

The Transformer architecture has become the de facto standard for Large Language Models (LLMs), demonstrating remarkable capabilities in language understanding and generation. However, its application in conversational AI is fundamentally constrained by its stateless nature and the quadratic computational complexity (O(L^2)) with respect to sequence length L. Current models emulate memory by reprocessing an ever-expanding conversation history with each turn, leading to prohibitive costs and latency in long dialogues. This paper introduces the Reactive Transformer (RxT), a novel architecture designed to overcome these limitations by shifting from a data-driven to an event-driven paradigm. RxT processes each conversational turn as a discrete event in real-time, maintaining context in an integrated, fixed-size Short-Term Memory (STM) system. The architecture features a distinct operational cycle where a generator-decoder produces a response based on the current query and the previous memory state, after which a memory-encoder and a dedicated Memory Attention network asynchronously update the STM with a representation of the complete interaction. This design fundamentally alters the scaling dynamics, reducing the total user-facing cost of a conversation from quadratic (O(N^2 cdot T)) to linear (O(N cdot T)) with respect to the number of interactions N. By decoupling response generation from memory updates, RxT achieves low latency, enabling truly real-time, stateful, and economically viable long-form conversations. We validated our architecture with a series of proof-of-concept experiments on synthetic data, demonstrating superior performance and constant-time inference latency compared to a baseline stateless model of comparable size.

The Transformer architecture, underpinned by the Multi-Head Attention (MHA) mechanism, has become the de facto standard for state-of-the-art models in artificial intelligence. However, the quadratic computational complexity of MHA with respect to sequence length presents a significant barrier to scaling, particularly for applications involving long contexts. Prevailing solutions, such as Multi-Query Attention (MQA) and Grouped-Query Attention (GQA), have effectively addressed the memory bandwidth bottleneck that dominates autoregressive inference latency by sharing Key and Value projections. While highly successful, these methods do not reduce the fundamental number of floating-point operations (FLOPs) required for the attention score computation, which remains a critical bottleneck for training and full-sequence processing. This paper introduces Sparse Query Attention (SQA), a novel attention architecture that pursues an alternative and complementary optimization path. Instead of reducing Key/Value heads, SQA reduces the number of Query heads. This architectural modification directly decreases the computational complexity of the attention mechanism by a factor proportional to the reduction in query heads, thereby lowering the overall FLOPs. This work presents the theoretical foundation of SQA, its mathematical formulation, and a family of architectural variants. Empirical benchmarks on long sequences (32k-200k tokens) demonstrate that SQA can achieve significant throughput improvements of up to 3x in computation-bound scenarios such as model pre-training, fine-tuning, and encoder-based tasks, with only a minimal impact on model quality in preliminary smallscale experiments. SQA was discovered serendipitously during the development of the upcoming Reactive Transformer architecture, suggesting its potential as a powerful tool for building more efficient and scalable models

Recent studies show that LLM hidden states encode reward-related information, such as answer correctness and model confidence. However, existing approaches typically fit black-box probes on the full hidden states, offering little insight into how this information is structured across neurons. In this paper, we show that reward-related information is concentrated in a sparse subset of neurons. Using simple probing, we identify two types of neurons: value neurons, whose activations predict state value, and dopamine neurons, whose activations encode step-level temporal difference (TD) errors. Together, these neurons form a sparse reward subsystem within LLM hidden states. These names are drawn by analogy with neuroscience, where value neurons and dopamine neurons in the biological reward subsystem also encode value and reward prediction errors, respectively. We demonstrate that value neurons are robust and transferable across diverse datasets and models, and provide causal evidence that they encode reward-related information. Finally, we show applications of the reward subsystem: value neurons serve as effective predictors of model confidence, and dopamine neurons can function as a process reward model (PRM) to guide inference-time search.

Large Language Models (LLMs) such as GPT-4o can handle a wide range of complex tasks with the right prompt. As per token costs are reduced, the advantages of fine-tuning Small Language Models (SLMs) for real-world applications -- faster inference, lower costs -- may no longer be clear. In this work, we present evidence that, for domain-specific tasks that require structured outputs, SLMs still have a quality advantage. We compare fine-tuning an SLM against prompting LLMs on the task of generating low-code workflows in JSON form. We observe that while a good prompt can yield reasonable results, fine-tuning improves quality by 10% on average. We also perform systematic error analysis to reveal model limitations.

Modern RL-based post-training for large language models (LLMs) co-locate trajectory sampling and policy optimisation on the same GPU cluster, forcing the system to switch between inference and training workloads. This serial context switching violates the single-program-multiple-data (SPMD) assumption underlying today's distributed training systems. We present Echo, the RL system that cleanly decouples these two phases across heterogeneous "inference" and "training" swarms while preserving statistical efficiency. Echo introduces two lightweight synchronization protocols: a sequential pull mode that refreshes policy weights according to API call for minimal bias, and an asynchronous push-pull mode that streams version-tagged rollouts through a replay buffer to maximise hardware utilisation. Training four representative RL workloads with Qwen3-4B, Qwen2.5-7B, Qwen3-30B-A3B-Thinking-2507 and Qwen3-32B on a geographically distributed cluster, Echo matches a fully co-located Verl baseline in convergence speed and final reward while off-loading trajectory generation to commodity edge hardware. These promising results demonstrate that large-scale RL for LLMs could achieve datacentre-grade performance using decentralised, heterogeneous resources.

Deploying a large language model (LLM) inference service remains costly because centralized serving depends on specialized GPU clusters and high-bandwidth interconnects in datacenters. An appealing alternative is to leverage collaborative decentralized GPU pools. However, heterogeneity in GPU and limited interconnected network bandwidth, along with potentially dynamic availability, make efficient scheduling the central challenge in this scenario. In this paper, we present Parallax, a decentralized LLM serving system that turns a pool of heterogeneous GPUs into an efficient inference platform via a two-phase scheduler. Parallax decomposes planning into (i) model allocation, which places layers of each replica across diverse GPUs to jointly optimize latency and throughput under memory and link-bandwidth constraints, and (ii) request-time GPU pipeline selection, which stitches layers from different replicas into end-to-end execution chains that balance load and adapt to current conditions. We implement Parallax and evaluate it on open-source LLMs deployed over real volunteer nodes. Parallax consistently reduces latency and increases throughput relative to decentralized baselines, demonstrating that principled scheduling can make volunteer compute a practical, affordable substrate for LLM inference. Github Repo at: https://github.com/GradientHQ/parallax.

Multi-agent systems (MAS) extend large language models (LLMs) from independent single-model reasoning to coordinative system-level intelligence. While existing LLM agents depend on text-based mediation for reasoning and communication, we take a step forward by enabling models to collaborate directly within the continuous latent space. We introduce LatentMAS, an end-to-end training-free framework that enables pure latent collaboration among LLM agents. In LatentMAS, each agent first performs auto-regressive latent thoughts generation through last-layer hidden embeddings. A shared latent working memory then preserves and transfers each agent's internal representations, ensuring lossless information exchange. We provide theoretical analyses establishing that LatentMAS attains higher expressiveness and lossless information preservation with substantially lower complexity than vanilla text-based MAS. In addition, empirical evaluations across 9 comprehensive benchmarks spanning math and science reasoning, commonsense understanding, and code generation show that LatentMAS consistently outperforms strong single-model and text-based MAS baselines, achieving up to 14.6% higher accuracy, reducing output token usage by 70.8%-83.7%, and providing 4x-4.3x faster end-to-end inference. These results demonstrate that our new latent collaboration framework enhances system-level reasoning quality while offering substantial efficiency gains without any additional training. Code and data are fully open-sourced at https://github.com/Gen-Verse/LatentMAS.

Retrieval-Augmented Generation enhances Large Language Models by integrating external knowledge, which reduces hallucinations but increases prompt length. This increase leads to higher computational costs and longer Time to First Token (TTFT). To mitigate this issue, existing solutions aim to reuse the preprocessed KV cache of each retrieved chunk to accelerate RAG. However, the lack of cross-chunk contextual information leads to a significant drop in generation quality, leaving the potential benefits of KV cache reuse largely unfulfilled. The challenge lies in how to reuse the precomputed KV cache of chunks while preserving generation quality. We propose FusionRAG, a novel inference framework that optimizes both the preprocessing and reprocessing stages of RAG. In the offline preprocessing stage, we embed information from other related text chunks into each chunk, while in the online reprocessing stage, we recompute the KV cache for tokens that the model focuses on. As a result, we achieve a better trade-off between generation quality and efficiency. According to our experiments, FusionRAG significantly improves generation quality at the same recomputation ratio compared to previous state-of-the-art solutions. By recomputing fewer than 15% of the tokens, FusionRAG achieves up to 70% higher normalized F1 scores than baselines and reduces TTFT by 2.66x-9.39x compared to Full Attention.

Most vehicle predictive maintenance systems rely exclusively on internal diagnostic signals and are validated on deterministic synthetic data, limiting the credibility of reported metrics. This paper presents a simulation-validated proof-of-concept framework for V2X-augmented predictive maintenance, integrating on-board sensor streams with external contextual signals -- road quality, weather, traffic density, and driver behaviour -- acquired via V2X communication and third-party APIs, with inference at the vehicle edge. Field validation on instrumented vehicles is identified as the required next step. Three experiments address common shortcomings of prior work. A feature group ablation study shows that V2X contextual features contribute a 2.6-point F1 gain, with full context removal reducing macro F1 from 0.855 to 0.807. On the AI4I 2020 real-world industrial failure dataset (10,000 samples, five failure modes), LightGBM achieves AUC-ROC of 0.973 under 5-fold stratified CV with SMOTE confined to training folds. A noise sensitivity analysis shows macro F1 remains above 0.88 under low noise and degrades to 0.74 under very high noise. SHAP analysis confirms that V2X and engineered interaction features rank among the top 15 predictors. Edge inference is estimated to reduce latency from 3.5s to under 1.0s versus cloud-only processing.

Recent advancements in large language models (LLMs) and their multimodal variants have led to remarkable progress across various domains, demonstrating impressive capabilities and unprecedented potential. In the era of ubiquitous connectivity, leveraging communication networks to distribute intelligence is a transformative concept, envisioning AI-powered services accessible at the network edge. However, pushing large models from the cloud to resource-constrained environments faces critical challenges. Model inference on low-end devices leads to excessive latency and performance bottlenecks, while raw data transmission over limited bandwidth networks causes high communication overhead. This article presents AI Flow, a framework that streamlines the inference process by jointly leveraging the heterogeneous resources available across devices, edge nodes, and cloud servers, making intelligence flow across networks. To facilitate cooperation among multiple computational nodes, the proposed framework explores a paradigm shift in the design of communication network systems from transmitting information flow to intelligence flow, where the goal of communications is task-oriented and folded into the inference process. Experimental results demonstrate the effectiveness of the proposed framework through an image captioning use case, showcasing the ability to reduce response latency while maintaining high-quality captions. This article serves as a position paper for identifying the motivation, challenges, and principles of AI Flow.

Reasoning and inference are central to human and artificial intelligence. Modeling inference in human language is very challenging. With the availability of large annotated data (Bowman et al., 2015), it has recently become feasible to train neural network based inference models, which have shown to be very effective. In this paper, we present a new state-of-the-art result, achieving the accuracy of 88.6% on the Stanford Natural Language Inference Dataset. Unlike the previous top models that use very complicated network architectures, we first demonstrate that carefully designing sequential inference models based on chain LSTMs can outperform all previous models. Based on this, we further show that by explicitly considering recursive architectures in both local inference modeling and inference composition, we achieve additional improvement. Particularly, incorporating syntactic parsing information contributes to our best result---it further improves the performance even when added to the already very strong model.

Large language models (LLMs) are widely applied in chatbots, code generators, and search engines. Workloads such as chain-of-thought, complex reasoning, and agent services significantly increase the inference cost by invoking the model repeatedly. Optimization methods such as parallelism, compression, and caching have been adopted to reduce costs, but the diverse service requirements make it hard to select the right method. Recently, specialized LLM inference engines have emerged as a key component for integrating the optimization methods into service-oriented infrastructures. However, a systematic study on inference engines is still lacking. This paper provides a comprehensive evaluation of 25 open-source and commercial inference engines. We examine each inference engine in terms of ease-of-use, ease-of-deployment, general-purpose support, scalability, and suitability for throughput- and latency-aware computation. Furthermore, we explore the design goals of each inference engine by investigating the optimization techniques it supports. In addition, we assess the ecosystem maturity of open source inference engines and handle the performance and cost policy of commercial solutions. We outline future research directions that include support for complex LLM-based services, support of various hardware, and enhanced security, offering practical guidance to researchers and developers in selecting and designing optimized LLM inference engines. We also provide a public repository to continually track developments in this fast-evolving field: https://github.com/sihyeong/Awesome-LLM-Inference-Engine

Goal-driven persuasive dialogue, exemplified by applications like telemarketing, requires sophisticated multi-turn planning and strict factual faithfulness, which remains a significant challenge for even state-of-the-art Large Language Models (LLMs). A lack of task-specific data often limits previous works, and direct LLM application suffers from strategic brittleness and factual hallucination. In this paper, we first construct and release TeleSalesCorpus, the first real-world-grounded dialogue dataset for this domain. We then propose AI-Salesman, a novel framework featuring a dual-stage architecture. For the training stage, we design a Bayesian-supervised reinforcement learning algorithm that learns robust sales strategies from noisy dialogues. For the inference stage, we introduce the Dynamic Outline-Guided Agent (DOGA), which leverages a pre-built script library to provide dynamic, turn-by-turn strategic guidance. Moreover, we design a comprehensive evaluation framework that combines fine-grained metrics for key sales skills with the LLM-as-a-Judge paradigm. Experimental results demonstrate that our proposed AI-Salesman significantly outperforms baseline models in both automatic metrics and comprehensive human evaluations, showcasing its effectiveness in complex persuasive scenarios.

In a striking neuroscience study, the authors placed a dead salmon in an MRI scanner and showed it images of humans in social situations. Astonishingly, standard analyses of the time reported brain regions predictive of social emotions. The explanation, of course, was not supernatural cognition but a cautionary tale about misapplied statistical inference. In AI interpretability, reports of similar ''dead salmon'' artifacts abound: feature attribution, probing, sparse auto-encoding, and even causal analyses can produce plausible-looking explanations for randomly initialized neural networks. In this work, we examine this phenomenon and argue for a pragmatic statistical-causal reframing: explanations of computational systems should be treated as parameters of a (statistical) model, inferred from computational traces. This perspective goes beyond simply measuring statistical variability of explanations due to finite sampling of input data; interpretability methods become statistical estimators, and findings should be tested against explicit and meaningful alternative computational hypotheses, with uncertainty quantified with respect to the postulated statistical model. It also highlights important theoretical issues, such as the identifiability of common interpretability queries, which we argue is critical to understand the field's susceptibility to false discoveries, poor generalizability, and high variance. More broadly, situating interpretability within the standard toolkit of statistical inference opens promising avenues for future work aimed at turning AI interpretability into a pragmatic and rigorous science.

AI agents have become increasingly adept at complex tasks such as coding, reasoning, and multimodal understanding. However, building generalist systems requires moving beyond individual agents to collective inference -- a paradigm where multi-agent systems with diverse, task-specialized agents complement one another through structured communication and collaboration. Today, coordination is usually handled with imprecise, ad-hoc natural language, which limits complex interaction and hinders interoperability with domain-specific agents. We introduce Agent context protocols (ACPs): a domain- and agent-agnostic family of structured protocols for agent-agent communication, coordination, and error handling. ACPs combine (i) persistent execution blueprints -- explicit dependency graphs that store intermediate agent outputs -- with (ii) standardized message schemas, enabling robust and fault-tolerant multi-agent collective inference. ACP-powered generalist systems reach state-of-the-art performance: 28.3 % accuracy on AssistantBench for long-horizon web assistance and best-in-class multimodal technical reports, outperforming commercial AI systems in human evaluation. ACPs are highly modular and extensible, allowing practitioners to build top-tier generalist agents quickly.

As the number of accepted papers at AI and ML conferences reaches into the thousands, it has become unclear how researchers access and read research publications. In this paper, we investigate the role of social media influencers in enhancing the visibility of machine learning research, particularly the citation counts of papers they share. We have compiled a comprehensive dataset of over 8,000 papers, spanning tweets from December 2018 to October 2023, alongside controls precisely matched by 9 key covariates. Our statistical and causal inference analysis reveals a significant increase in citations for papers endorsed by these influencers, with median citation counts 2-3 times higher than those of the control group. Additionally, the study delves into the geographic, gender, and institutional diversity of highlighted authors. Given these findings, we advocate for a responsible approach to curation, encouraging influencers to uphold the journalistic standard that includes showcasing diverse research topics, authors, and institutions.

Large language models are now embedded in everyday writing workflows, making reliable AI-generated text detection important for academic integrity, content moderation, and provenance tracking. In practice, however, a detector must do more than achieve high aggregate AUROC on clean, in-distribution human and AI text: it should remain robust to attacks and adversarial rewrites, transfer to unseen generators and domains, and operate at low false-positive rates (FPR). Most existing detectors optimize a single AI/Human objective, giving the representation little incentive to learn generator, attack, or domain structure once the binary task saturates. We introduce MELD (Multi-Task Equilibrated Learning Detector), a deployable detector for AI-generated text that enriches binary detection with auxiliary supervision. MELD attaches generator-family, attack-type, and source-domain heads to a shared encoder, and balances the four losses with learned homoscedastic uncertainty weights. To improve robustness, an EMA teacher predicts on clean inputs while an attack-augmented student is distilled toward the teacher. MELD further uses a hard-negative pairwise ranking loss to enlarge the score margin between AI-generated texts and the most confusable human texts. At inference, all auxiliary heads are discarded, giving MELD the same interface and cost as a standard detector. On the public RAID leaderboard, MELD is the strongest open-source detector and is competitive with leading commercial models, especially under attack and at low FPR. Across standard held-out benchmarks, MELD matches or outperforms supervised baselines. We further introduce MELD-eval, a held-out evaluation pool built from recent chat models released by four major LLM providers. Without additional finetuning, MELD achieves 99.9% TPR at 1% FPR on MELD-eval, while many baselines degrade sharply.

EigenAI is a verifiable AI platform built on top of the EigenLayer restaking ecosystem. At a high level, it combines a deterministic large-language model (LLM) inference engine with a cryptoeconomically secured optimistic re-execution protocol so that every inference result can be publicly audited, reproduced, and, if necessary, economically enforced. An untrusted operator runs inference on a fixed GPU architecture, signs and encrypts the request and response, and publishes the encrypted log to EigenDA. During a challenge window, any watcher may request re-execution through EigenVerify; the result is then deterministically recomputed inside a trusted execution environment (TEE) with a threshold-released decryption key, allowing a public challenge with private data. Because inference itself is bit-exact, verification reduces to a byte-equality check, and a single honest replica suffices to detect fraud. We show how this architecture yields sovereign agents -- prediction-market judges, trading bots, and scientific assistants -- that enjoy state-of-the-art performance while inheriting security from Ethereum's validator base.

AI-generated text is proliferating across domains, from creative writing and journalism to marketing content and scientific articles. Models can follow user-provided instructions to generate coherent and grammatically correct outputs but in this work, we study a more fundamental question: how do we evaluate and improve the writing quality of AI-generated text? Writing quality assessment has received less attention from the community, in part because it is fundamentally subjective and requires expertise. We first introduce the Writing Quality Benchmark (WQ) by consolidating five writing-preference datasets into 4,729 writing quality judgments. Our experiments show that most of the competitive baselines, including state-of-the-art LLMs that excel at reasoning tasks, barely outperform random baselines on WQ. We then train specialized Writing Quality Reward Models (WQRM) of various sizes for writing quality assessment that demonstrate strong generalization on four out-of-distribution test sets and 74% accuracy on the WQ benchmark. To further show WQRM's practical benefits during inference, we leverage additional test-time compute to generate and rank multiple candidate revisions, allowing us to select higher-quality outputs from an initial draft. Human evaluation with 9 experienced writers confirm that WQRM-based selection produces writing samples preferred by experts 66% overall, and 72.2% when the reward gap is larger than 1 point. We release our datasets and models to encourage community engagement with writing quality assessment and development of AI writing systems better aligned with human preferences.

Unlike most reinforcement learning agents which require an unrealistic amount of environment interactions to learn a new behaviour, humans excel at learning quickly by merely observing and imitating others. This ability highly depends on the fact that humans have a model of their own embodiment that allows them to infer the most likely actions that led to the observed behaviour. In this paper, we propose Action Inference by Maximising Evidence (AIME) to replicate this behaviour using world models. AIME consists of two distinct phases. In the first phase, the agent learns a world model from its past experience to understand its own body by maximising the ELBO. While in the second phase, the agent is given some observation-only demonstrations of an expert performing a novel task and tries to imitate the expert's behaviour. AIME achieves this by defining a policy as an inference model and maximising the evidence of the demonstration under the policy and world model. Our method is "zero-shot" in the sense that it does not require further training for the world model or online interactions with the environment after given the demonstration. We empirically validate the zero-shot imitation performance of our method on the Walker and Cheetah embodiment of the DeepMind Control Suite and find it outperforms the state-of-the-art baselines. Code is available at: https://github.com/argmax-ai/aime.

Reward modelling from preference data is a crucial step in aligning large language models (LLMs) with human values, requiring robust generalisation to novel prompt-response pairs. In this work, we propose to frame this problem in a causal paradigm, providing the rich toolbox of causality to identify the persistent challenges, such as causal misidentification, preference heterogeneity, and confounding due to user-specific factors. Inheriting from the literature of causal inference, we identify key assumptions necessary for reliable generalisation and contrast them with common data collection practices. We illustrate failure modes of naive reward models and demonstrate how causally-inspired approaches can improve model robustness. Finally, we outline desiderata for future research and practices, advocating targeted interventions to address inherent limitations of observational data.

Emerging AI accelerators increasingly adopt wafer-scale manufacturing technologies, integrating hundreds of thousands of AI cores in a mesh architecture with large distributed on-chip memory (tens of GB in total) and ultra-high on-chip memory bandwidth (tens of PB/s). However, current LLM inference systems, optimized for shared memory architectures like GPUs, fail to exploit these accelerators fully. We introduce WaferLLM, the first wafer-scale LLM inference system. WaferLLM is guided by a novel PLMR model (pronounced as "Plummer") that captures the unique hardware characteristics of wafer-scale architectures. Leveraging this model, WaferLLM pioneers wafer-scale LLM parallelism, optimizing the utilization of hundreds of thousands of on-chip cores. It also introduces MeshGEMM and MeshGEMV, the first GEMM and GEMV implementations designed to scale effectively on wafer-scale accelerators. Evaluations show that WaferLLM achieves up to 200times higher accelerator utilization than state-of-the-art methods. Leveraging a wafer-scale accelerator (Cerebras WSE2), WaferLLM delivers GEMV operations 606times faster and 16times more energy-efficient than on an NVIDIA A100 GPU. For full LLM inference, WaferLLM achieves 10-20times speedups over A100 GPU clusters running SGLang and vLLM. These advantages are expected to grow as wafer-scale AI models, software, and hardware continue to mature. WaferLLM is open-sourced at https://github.com/MeshInfra/WaferLLM.

Large transformer models display promising performance on a wide range of natural language processing (NLP) tasks. Although the AI community has expanded the model scale to the trillion parameter level, the practical deployment of 10-100 billion parameter models is still uncertain due to the latency, throughput, and memory constraints. In this paper, we proposed EnergonAI to solve the challenges of the efficient deployment of 10-100 billion parameter transformer models on single- or multi-GPU systems. EnergonAI adopts a hierarchy-controller system architecture to coordinate multiple devices and efficiently support different parallel patterns. It delegates the execution of sub-models to multiple workers in the single-controller style and applies tensor parallelism and pipeline parallelism among the workers in a multi-controller style. Upon the novel architecture, we propose three techniques, i.e. non-blocking pipeline parallelism, distributed redundant computation elimination, and peer memory pooling. EnergonAI enables the users to program complex parallel code the same as a serial one. Compared with the FasterTransformer, we have proven that EnergonAI has superior performance on latency and throughput. In our experiments, EnergonAI can achieve 37% latency reduction in tensor parallelism, 10% scalability improvement in pipeline parallelism, and it improves the model scale inferred on a single GPU by using a larger heterogeneous memory space at cost of limited performance reduction.

Explanatory inference is the creation and evaluation of hypotheses that provide explanations, and is sometimes known as abduction or abductive inference. Generative AI is a new set of artificial intelligence models based on novel algorithms for generating text, images, and sounds. This paper proposes a set of benchmarks for assessing the ability of AI programs to perform explanatory inference, and uses them to determine the extent to which ChatGPT, a leading generative AI model, is capable of making explanatory inferences. Tests on the benchmarks reveal that ChatGPT performs creative and evaluative inferences in many domains, although it is limited to verbal and visual modalities. Claims that ChatGPT and similar models are incapable of explanation, understanding, causal reasoning, meaning, and creativity are rebutted.

Monolithic large language models (LLMs) like GPT-4 have paved the way for modern generative AI applications. Training, serving, and maintaining monolithic LLMs at scale, however, remains prohibitively expensive and challenging. The disproportionate increase in compute-to-memory ratio of modern AI accelerators have created a memory wall, necessitating new methods to deploy AI. Composition of Experts (CoE) is an alternative modular approach that lowers the cost and complexity of training and serving. However, this approach presents two key challenges when using conventional hardware: (1) without fused operations, smaller models have lower operational intensity, which makes high utilization more challenging to achieve; and (2) hosting a large number of models can be either prohibitively expensive or slow when dynamically switching between them. In this paper, we describe how combining CoE, streaming dataflow, and a three-tier memory system scales the AI memory wall. We describe Samba-CoE, a CoE system with 150 experts and a trillion total parameters. We deploy Samba-CoE on the SambaNova SN40L Reconfigurable Dataflow Unit (RDU) - a commercial dataflow accelerator architecture that has been co-designed for enterprise inference and training applications. The chip introduces a new three-tier memory system with on-chip distributed SRAM, on-package HBM, and off-package DDR DRAM. A dedicated inter-RDU network enables scaling up and out over multiple sockets. We demonstrate speedups ranging from 2x to 13x on various benchmarks running on eight RDU sockets compared with an unfused baseline. We show that for CoE inference deployments, the 8-socket RDU Node reduces machine footprint by up to 19x, speeds up model switching time by 15x to 31x, and achieves an overall speedup of 3.7x over a DGX H100 and 6.6x over a DGX A100.

Precision therapeutics require multimodal adaptive models that generate personalized treatment recommendations. We introduce TxAgent, an AI agent that leverages multi-step reasoning and real-time biomedical knowledge retrieval across a toolbox of 211 tools to analyze drug interactions, contraindications, and patient-specific treatment strategies. TxAgent evaluates how drugs interact at molecular, pharmacokinetic, and clinical levels, identifies contraindications based on patient comorbidities and concurrent medications, and tailors treatment strategies to individual patient characteristics. It retrieves and synthesizes evidence from multiple biomedical sources, assesses interactions between drugs and patient conditions, and refines treatment recommendations through iterative reasoning. It selects tools based on task objectives and executes structured function calls to solve therapeutic tasks that require clinical reasoning and cross-source validation. The ToolUniverse consolidates 211 tools from trusted sources, including all US FDA-approved drugs since 1939 and validated clinical insights from Open Targets. TxAgent outperforms leading LLMs, tool-use models, and reasoning agents across five new benchmarks: DrugPC, BrandPC, GenericPC, TreatmentPC, and DescriptionPC, covering 3,168 drug reasoning tasks and 456 personalized treatment scenarios. It achieves 92.1% accuracy in open-ended drug reasoning tasks, surpassing GPT-4o and outperforming DeepSeek-R1 (671B) in structured multi-step reasoning. TxAgent generalizes across drug name variants and descriptions. By integrating multi-step inference, real-time knowledge grounding, and tool-assisted decision-making, TxAgent ensures that treatment recommendations align with established clinical guidelines and real-world evidence, reducing the risk of adverse events and improving therapeutic decision-making.

Large language model (LLM) queries are predominantly processed by frontier models in centralized cloud infrastructure. Rapidly growing demand strains this paradigm, and cloud providers struggle to scale infrastructure at pace. Two advances enable us to rethink this paradigm: small LMs (<=20B active parameters) now achieve competitive performance to frontier models on many tasks, and local accelerators (e.g., Apple M4 Max) run these models at interactive latencies. This raises the question: can local inference viably redistribute demand from centralized infrastructure? Answering this requires measuring whether local LMs can accurately answer real-world queries and whether they can do so efficiently enough to be practical on power-constrained devices (i.e., laptops). We propose intelligence per watt (IPW), task accuracy divided by unit of power, as a metric for assessing capability and efficiency of local inference across model-accelerator pairs. We conduct a large-scale empirical study across 20+ state-of-the-art local LMs, 8 accelerators, and a representative subset of LLM traffic: 1M real-world single-turn chat and reasoning queries. For each query, we measure accuracy, energy, latency, and power. Our analysis reveals 3 findings. First, local LMs can accurately answer 88.7% of single-turn chat and reasoning queries with accuracy varying by domain. Second, from 2023-2025, IPW improved 5.3x and local query coverage rose from 23.2% to 71.3%. Third, local accelerators achieve at least 1.4x lower IPW than cloud accelerators running identical models, revealing significant headroom for optimization. These findings demonstrate that local inference can meaningfully redistribute demand from centralized infrastructure, with IPW serving as the critical metric for tracking this transition. We release our IPW profiling harness for systematic intelligence-per-watt benchmarking.

The surge in generative AI workloads has created a need for scalable inference systems that can flexibly harness both GPUs and specialized accelerators while containing operational costs. This paper proposes a hardware-agnostic control loop that adaptively allocates requests across heterogeneous accelerators based on real-time cost and capacity signals. The approach sustains low latency and high throughput by dynamically shifting between cost-optimized and capacity-optimized modes, ensuring the most efficient use of expensive compute resources under fluctuating availability. Evaluated using the Stable Diffusion model, the framework consistently meets latency targets, automatically redirects traffic during capacity shortfalls, and capitalizes on lower-cost accelerators when possible. These results highlight how a feedback-driven deployment strategy, spanning the entire software and hardware stack, can help organizations efficiently scale generative AI workloads while maintaining resilience in the face of limited accelerator capacity.

The rapid development of AI-generated content (AIGC) technology has led to the misuse of highly realistic AI-generated images (AIGI) in spreading misinformation, posing a threat to public information security. Although existing AIGI detection techniques are generally effective, they face two issues: 1) a lack of human-verifiable explanations, and 2) a lack of generalization in the latest generation technology. To address these issues, we introduce a large-scale and comprehensive dataset, Holmes-Set, which includes the Holmes-SFTSet, an instruction-tuning dataset with explanations on whether images are AI-generated, and the Holmes-DPOSet, a human-aligned preference dataset. Our work introduces an efficient data annotation method called the Multi-Expert Jury, enhancing data generation through structured MLLM explanations and quality control via cross-model evaluation, expert defect filtering, and human preference modification. In addition, we propose Holmes Pipeline, a meticulously designed three-stage training framework comprising visual expert pre-training, supervised fine-tuning, and direct preference optimization. Holmes Pipeline adapts multimodal large language models (MLLMs) for AIGI detection while generating human-verifiable and human-aligned explanations, ultimately yielding our model AIGI-Holmes. During the inference stage, we introduce a collaborative decoding strategy that integrates the model perception of the visual expert with the semantic reasoning of MLLMs, further enhancing the generalization capabilities. Extensive experiments on three benchmarks validate the effectiveness of our AIGI-Holmes.

Enabling agentic AI systems to adapt their problem-solving approaches based on post-training interactions remains a fundamental challenge. While systems that update and maintain a memory at inference time have been proposed, existing designs only steer the system by modifying textual input to a language model or agent, which means that they cannot change sampling parameters, remove tools, modify system prompts, or switch between agentic and workflow paradigms. On the other hand, systems that adapt more flexibly require offline optimization and remain static once deployed. We present Experience-Guided Reasoner (EGuR), which generates tailored strategies -- complete computational procedures involving LLM calls, tools, sampling parameters, and control logic -- dynamically at inference time based on accumulated experience. We achieve this using an LLM-based meta-strategy -- a strategy that outputs strategies -- enabling adaptation of all strategy components (prompts, sampling parameters, tool configurations, and control logic). EGuR operates through two components: a Guide generates multiple candidate strategies conditioned on the current problem and structured memory of past experiences, while a Consolidator integrates execution feedback to improve future strategy generation. This produces complete, ready-to-run strategies optimized for each problem, which can be cached, retrieved, and executed as needed without wasting resources. Across five challenging benchmarks (AIME 2025, 3-SAT, and three Big Bench Extra Hard tasks), EGuR achieves up to 14% accuracy improvements over the strongest baselines while reducing computational costs by up to 111x, with both metrics improving as the system gains experience.

The real-time deployment of cascaded generative AI pipelines for applications like video translation is constrained by significant system-level challenges. These include the cumulative latency of sequential model inference and the quadratic (O(N^2)) computational complexity that renders multi-user video conferencing applications unscalable. This paper proposes and evaluates a practical system-level framework designed to mitigate these critical bottlenecks. The proposed architecture incorporates a turn-taking mechanism to reduce computational complexity from quadratic to linear in multi-user scenarios, and a segmented processing protocol to manage inference latency for a perceptually real-time experience. We implement a proof-of-concept pipeline and conduct a rigorous performance analysis across a multi-tiered hardware setup, including commodity (NVIDIA RTX 4060), cloud (NVIDIA T4), and enterprise (NVIDIA A100) GPUs. Our objective evaluation demonstrates that the system achieves real-time throughput (τ< 1.0) on modern hardware. A subjective user study further validates the approach, showing that a predictable, initial processing delay is highly acceptable to users in exchange for a smooth, uninterrupted playback experience. The work presents a validated, end-to-end system design that offers a practical roadmap for deploying scalable, real-time generative AI applications in multilingual communication platforms.

Agentic workflows, where multiple AI agents collaborate to accomplish complex tasks like reasoning or planning, are becoming increasingly prevalent. However, these workflows often suffer from error propagation and sub-optimal performance, largely due to poorly designed prompts that fail to effectively guide individual agents. This is a critical problem because it limits the reliability and scalability of these powerful systems. We introduce ProRefine, an innovative inference-time prompt optimization method that leverages textual feedback from large language models (LLMs) to address this challenge. ProRefine dynamically refines prompts for multi-step reasoning tasks without additional training or ground truth labels. Evaluated on five benchmark mathematical reasoning datasets, ProRefine significantly surpasses zero-shot Chain-of-Thought baselines by 3 to 37 percentage points. This approach not only boosts accuracy but also allows smaller models to match the performance of larger ones, highlighting its potential for efficient and scalable AI deployment, and democratizing access to high-performing AI.

Low-precision data types are essential in modern neural networks during both training and inference as they enhance throughput and computational capacity by better exploiting available hardware resources. Despite the incorporation of FP8 in commercially available neural network accelerators, a comprehensive exposition of its underlying mechanisms, along with rigorous performance and accuracy evaluations, is still lacking. In this work, we contribute in three significant ways. First, we analyze the implementation details and quantization options associated with FP8 for inference on the Intel Gaudi AI accelerator. Second, we empirically quantify the throughput improvements afforded by the use of FP8 at both the operator level and in end-to-end scenarios. Third, we assess the accuracy impact of various FP8 quantization methods. Our experimental results indicate that the Intel Gaudi 2 accelerator consistently achieves high computational unit utilization, frequently exceeding 90% MFU, while incurring an accuracy degradation of less than 1%.

Consider a market of competing model providers selling query access to models with varying costs and capabilities. Customers submit problem instances and are willing to pay up to a budget for a verifiable solution. An arbitrageur efficiently allocates inference budget across providers to undercut the market, thus creating a competitive offering with no model-development risk. In this work, we initiate the study of arbitrage in AI model markets, empirically demonstrating the viability of arbitrage and illustrating its economic consequences. We conduct an in-depth case study of SWE-bench GitHub issue resolution using two representative models, GPT-5 mini and DeepSeek v3.2. In this verifiable domain, simple arbitrage strategies generate net profit margins of up to 40%. Robust arbitrage strategies that generalize across different domains remain profitable. Distillation further creates strong arbitrage opportunities, potentially at the expense of the teacher model's revenue. Multiple competing arbitrageurs drive down consumer prices, reducing the marginal revenue of model providers. At the same time, arbitrage reduces market segmentation and facilitates market entry for smaller model providers by enabling earlier revenue capture. Our results suggest that arbitrage can be a powerful force in AI model markets with implications for model development, distillation, and deployment.

Personal Artificial Intelligence is currently hindered by the fragmentation of user data across isolated silos. While Retrieval-Augmented Generation offers a partial remedy, its reliance on unstructured vector similarity fails to capture the latent semantic topology and temporal dependencies essential for holistic sensemaking. We introduce EpisTwin, a neuro-symbolic framework that grounds generative reasoning in a verifiable, user-centric Personal Knowledge Graph. EpisTwin leverages Multimodal Language Models to lift heterogeneous, cross-application data into semantic triples. At inference, EpisTwin enables complex reasoning over the personal semantic graph via an agentic coordinator that combines Graph Retrieval-Augmented Generation with Online Deep Visual Refinement, dynamically re-grounding symbolic entities in their raw visual context. We also introduce PersonalQA-71-100, a synthetic benchmark designed to simulate a realistic user's digital footprint and evaluate EpisTwin performance. Our framework demonstrates robust results across a suite of state-of-the-art judge models, offering a promising direction for trustworthy Personal AI.

Large language models have become central to many AI applications, but their growing energy consumption raises serious sustainability concerns. A key limitation in current AI deployments is the reliance on a one-size-fits-all inference strategy where most systems route every request to the same large model, regardless of task complexity, leading to substantial and unnecessary energy waste. To address this issue, we propose a context-aware model switching approach that dynamically selects an appropriate language model based on query complexity. The proposed system uses a Context-Aware Model Switching for Energy-Efficient LLM Inference that combines caching for repeated queries, rulebased complexity scoring for fast and explainable decisions, machine learning classification to capture semantic intent, and a user-adaptive component that learns from interaction patterns over time. The proposed architecture was evaluated using real conversation workloads and three open-source language models (Gemma3 1B, Gemma3 4B and Qwen3 4B) with different computational costs, measuring energy consumption (via NVML GPU power telemetry), response latency, routing accuracy, and output quality (BERTScore F1) to reflect real-world usage conditions. Experimental results show that the model switching approach can reduce energy consumption by up to 67.5% compared to always using the largest model while maintaining a response quality of 93.6%. In addition, the response time for simple queries also improved significantly by approximately 68%. These results show that model switching inference offers a practical and scalable path toward more energy-efficient and sustainable AI systems, demonstrating that significant efficiency gains can be achieved without major sacrifices in response quality.

The spreading of AI-generated images (AIGI), driven by advances in generative AI, poses a significant threat to information security and public trust. Existing AIGI detectors, while effective against images in clean laboratory settings, fail to generalize to in-the-wild scenarios. These real-world images are noisy, varying from ``obviously fake" images to realistic ones derived from multiple generative models and further edited for quality control. We address in-the-wild AIGI detection in this paper. We introduce Mirage, a challenging benchmark designed to emulate the complexity of in-the-wild AIGI. Mirage is constructed from two sources: (1) a large corpus of Internet-sourced AIGI verified by human experts, and (2) a synthesized dataset created through the collaboration between multiple expert generators, closely simulating the realistic AIGI in the wild. Building on this benchmark, we propose Mirage-R1, a vision-language model with heuristic-to-analytic reasoning, a reflective reasoning mechanism for AIGI detection. Mirage-R1 is trained in two stages: a supervised-fine-tuning cold start, followed by a reinforcement learning stage. By further adopting an inference-time adaptive thinking strategy, Mirage-R1 is able to provide either a quick judgment or a more robust and accurate conclusion, effectively balancing inference speed and performance. Extensive experiments show that our model leads state-of-the-art detectors by 5% and 10% on Mirage and the public benchmark, respectively. The benchmark and code will be made publicly available.

AI agents are emerging as a dominant workload in a wide range of applications, promising to be the vehicle that delivers the promised benefits of AI to enterprises and consumers. Unlike conventional software or static inference, agentic workloads are dynamic and structurally complex. Often these agents are directed graphs of compute and IO operations that span multi-modal data input and conversion), data processing and context gathering (e.g vector DB lookups), multiple LLM inferences, tool calls, etc. To scale AI agent usage, we need efficient and scalable deployment and agent-serving infrastructure. To tackle this challenge, in this paper, we present a system design for dynamic orchestration of AI agent workloads on heterogeneous compute infrastructure spanning CPUs and accelerators, both from different vendors and across different performance tiers within a single vendor. The system delivers several building blocks: a framework for planning and optimizing agentic AI execution graphs using cost models that account for compute, memory, and bandwidth constraints of different HW; a MLIR based representation and compilation system that can decompose AI agent execution graphs into granular operators and generate code for different HW options; and a dynamic orchestration system that can place the granular components across a heterogeneous compute infrastructure and stitch them together while meeting an end-to-end SLA. Our design performs a systems level TCO optimization and preliminary results show that leveraging a heterogeneous infrastructure can deliver significant TCO benefits. A preliminary surprising finding is that for some workloads a heterogeneous combination of older generation GPUs with newer accelerators can deliver similar TCO as the latest generation homogenous GPU infrastructure design, potentially extending the life of deployed infrastructure.

Despite advancements in causal inference and prescriptive AI, its adoption in enterprise settings remains hindered primarily due to its technical complexity. Many users lack the necessary knowledge and appropriate tools to effectively leverage these technologies. This work at the MIT-IBM Watson AI Lab focuses on developing the proof-of-concept agent, PrecAIse, a domain-adaptable conversational agent equipped with a suite of causal and prescriptive tools to help enterprise users make better business decisions. The objective is to make advanced, novel causal inference and prescriptive tools widely accessible through natural language interactions. The presented Natural Language User Interface (NLUI) enables users with limited expertise in machine learning and data science to harness prescriptive analytics in their decision-making processes without requiring intensive computing resources. We present an agent capable of function calling, maintaining faithful, interactive, and dynamic conversations, and supporting new domains.

In recent years, large language models have demonstrated remarkable performance across various natural language processing (NLP) tasks. However, deploying these models for real-world applications often requires efficient inference solutions to handle the computational demands. In this paper, we explore the utilization of CPUs for accelerating the inference of large language models. Specifically, we introduce a parallelized approach to enhance throughput by 1) Exploiting the parallel processing capabilities of modern CPU architectures, 2) Batching the inference request. Our evaluation shows the accelerated inference engine gives an 18-22x improvement in the generated token per sec. The improvement is more with longer sequence and larger models. In addition to this, we can also run multiple workers in the same machine with NUMA node isolation to further improvement in tokens/s. Table 2, we have received 4x additional improvement with 4 workers. This would also make Gen-AI based products and companies environment friendly, our estimates shows that CPU usage for Inference could reduce the power consumption of LLMs by 48.9% while providing production ready throughput and latency.

Recent developments in Large Language Models (LLMs) have shifted from pre-training scaling to post-training and test-time scaling. Across these developments, a key unified paradigm has arisen: Learning from Rewards, where reward signals act as the guiding stars to steer LLM behavior. It has underpinned a wide range of prevalent techniques, such as reinforcement learning (in RLHF, DPO, and GRPO), reward-guided decoding, and post-hoc correction. Crucially, this paradigm enables the transition from passive learning from static data to active learning from dynamic feedback. This endows LLMs with aligned preferences and deep reasoning capabilities. In this survey, we present a comprehensive overview of the paradigm of learning from rewards. We categorize and analyze the strategies under this paradigm across training, inference, and post-inference stages. We further discuss the benchmarks for reward models and the primary applications. Finally we highlight the challenges and future directions. We maintain a paper collection at https://github.com/bobxwu/learning-from-rewards-llm-papers.

Large Language Model (LLM) inference on large-scale systems is expected to dominate future cloud infrastructures. Efficient LLM inference in cloud environments with numerous AI accelerators is challenging, necessitating extensive optimizations for optimal performance. Current systems batch prefill and decoding to boost throughput but encounter latency issues, while others disaggregate these phases, leading to resource underutilization. We propose AcceLLM, a novel method addressing latency and load balancing, inspired by the cache data management. It strategically utilizes redundant data to enhance inference via load balancing and optimal hardware use. Simulated evaluations on Nvidia H100 GPU and Huawei Ascend 910B2 show AcceLLM surpasses state-of-the-art systems up to 30% in latency and efficiency, handling diverse workloads effectively.

Large Language Model (LLM) inference is hard. The autoregressive Decode phase of the underlying Transformer model makes LLM inference fundamentally different from training. Exacerbated by recent AI trends, the primary challenges are memory and interconnect rather than compute. To address these challenges, we highlight four architecture research opportunities: High Bandwidth Flash for 10X memory capacity with HBM-like bandwidth; Processing-Near-Memory and 3D memory-logic stacking for high memory bandwidth; and low-latency interconnect to speedup communication. While our focus is datacenter AI, we also review their applicability for mobile devices.

As the adoption of Generative AI in real-world services grow explosively, energy has emerged as a critical bottleneck resource. However, energy remains a metric that is often overlooked, under-explored, or poorly understood in the context of building ML systems. We present the ML.ENERGY Benchmark, a benchmark suite and tool for measuring inference energy consumption under realistic service environments, and the corresponding ML.ENERGY Leaderboard, which have served as a valuable resource for those hoping to understand and optimize the energy consumption of their generative AI services. In this paper, we explain four key design principles for benchmarking ML energy we have acquired over time, and then describe how they are implemented in the ML.ENERGY Benchmark. We then highlight results from the latest iteration of the benchmark, including energy measurements of 40 widely used model architectures across 6 different tasks, case studies of how ML design choices impact energy consumption, and how automated optimization recommendations can lead to significant (sometimes more than 40%) energy savings without changing what is being computed by the model. The ML.ENERGY Benchmark is open-source and can be easily extended to various customized models and application scenarios.

Large language models (LLMs) like GPTs, trained on vast datasets, have demonstrated impressive capabilities in language understanding, reasoning, and planning, achieving human-level performance in various tasks. Most studies focus on enhancing these models by training on ever-larger datasets to build more powerful foundation models. While training stronger models is important, enabling models to evolve during inference is equally crucial, a process we refer to as AI self-evolution. Unlike large-scale training, self-evolution may rely on limited data or interactions. Inspired by the columnar organization of the human cerebral cortex, we hypothesize that AI models could develop cognitive abilities and build internal representations through iterative interactions with their environment. To achieve this, models need long-term memory (LTM) to store and manage processed interaction data. LTM supports self-evolution by representing diverse experiences across environments and agents. In this report, we explore AI self-evolution and its potential to enhance models during inference. We examine LTM's role in lifelong learning, allowing models to evolve based on accumulated interactions. We outline the structure of LTM and the systems needed for effective data retention and representation. We also classify approaches for building personalized models with LTM data and show how these models achieve self-evolution through interaction. Using LTM, our multi-agent framework OMNE achieved first place on the GAIA benchmark, demonstrating LTM's potential for AI self-evolution. Finally, we present a roadmap for future research, emphasizing the importance of LTM for advancing AI technology and its practical applications.

In current AI era, users may request AI companies to delete their data from the training dataset due to the privacy concerns. As a model owner, retraining a model will consume significant computational resources. Therefore, machine unlearning is a new emerged technology to allow model owner to delete requested training data or a class with little affecting on the model performance. However, for large-scaling complex data, such as image or text data, unlearning a class from a model leads to a inferior performance due to the difficulty to identify the link between classes and model. An inaccurate class deleting may lead to over or under unlearning. In this paper, to accurately defining the unlearning class of complex data, we apply the definition of Concept, rather than an image feature or a token of text data, to represent the semantic information of unlearning class. This new representation can cut the link between the model and the class, leading to a complete erasing of the impact of a class. To analyze the impact of the concept of complex data, we adopt a Post-hoc Concept Bottleneck Model, and Integrated Gradients to precisely identify concepts across different classes. Next, we take advantage of data poisoning with random and targeted labels to propose unlearning methods. We test our methods on both image classification models and large language models (LLMs). The results consistently show that the proposed methods can accurately erase targeted information from models and can largely maintain the performance of the models.

Large AI models (e.g., Dall-E, GPT4) have electrified the scientific, technological and societal landscape through their superhuman capabilities. These services are offered largely in a traditional web2.0 format (e.g., OpenAI's GPT4 service). As more large AI models proliferate (personalizing and specializing to a variety of domains), there is a tremendous need to have a neutral trust-free platform that allows the hosting of AI models, clients receiving AI services efficiently, yet in a trust-free, incentive compatible, Byzantine behavior resistant manner. In this paper we propose SAKSHI, a trust-free decentralized platform specifically suited for AI services. The key design principles of SAKSHI are the separation of the data path (where AI query and service is managed) and the control path (where routers and compute and storage hosts are managed) from the transaction path (where the metering and billing of services are managed over a blockchain). This separation is enabled by a "proof of inference" layer which provides cryptographic resistance against a variety of misbehaviors, including poor AI service, nonpayment for service, copying of AI models. This is joint work between multiple universities (Princeton University, University of Illinois at Urbana-Champaign, Tsinghua University, HKUST) and two startup companies (Witness Chain and Eigen Layer).

Large language models (LLMs) power a new generation of interactive AI applications exemplified by ChatGPT. The interactive nature of these applications demand low job completion time (JCT) for model inference. Existing LLM serving systems use run-to-completion processing for inference jobs, which suffers from head-of-line blocking and long JCT. We present FastServe, a distributed inference serving system for LLMs. FastServe exploits the autoregressive pattern of LLM inference to enable preemption at the granularity of each output token. FastServe uses preemptive scheduling to minimize JCT with a novel skip-join Multi-Level Feedback Queue scheduler. Based on the new semi information-agnostic setting of LLM inference, the scheduler leverages the input length information to assign an appropriate initial queue for each arrival job to join. The higher priority queues than the joined queue are skipped to reduce demotions. We design an efficient GPU memory management mechanism that proactively offloads and uploads intermediate states between GPU memory and host memory for LLM inference. We build a system prototype of FastServe based on NVIDIA FasterTransformer. Experimental results show that compared to the state-of-the-art solution Orca, FastServe improves the average and tail JCT by up to 5.1times and 6.4times, respectively.

Legal reasoning is not semantic similarity search. A court judgment encodes constrained symbolic reasoning: precedent propagation, procedural state transitions, and statute-bound inference. These are properties that vector-based retrieval-augmented generation (RAG) cannot faithfully represent. Hallucinated precedents, outdated statute citations, and unsupported reasoning chains remain persistent failure modes in LLM-based legal AI, with real consequences for access to justice in high-caseload jurisdictions such as India. This paper presents Falkor-IRAC, a graph-constrained generation framework for Indian legal AI that grounds generation in structured reasoning over an IRAC (Issue, Rule, Analysis, Conclusion) knowledge graph. Judgments from the Supreme Court and High Courts of India are ingested as IRAC node structures enriched with procedural state transitions, precedent relationships, and statutory references, stored in FalkorDB for low-latency agentic traversal. At inference time, LLM-generated answers are accepted only if a valid supporting path can be traced through the graph, a check performed by a falsifiability oracle called the Verifier Agent. The system also detects doctrinal conflicts as a first-class output rather than silently resolving them. Falkor-IRAC is evaluated using graph-native metrics: citation grounding accuracy, path validity rate, hallucinated precedent rate, and conflict detection rate. These metrics are argued to be more appropriate for legal reasoning evaluation than BLEU and ROUGE. On a proof-of-concept corpus of 51 Supreme Court judgments, the Verifier Agent correctly validated citations on completed queries and correctly rejected fabricated citations. Evaluation against vector-only RAG baselines is left for future work, as is GPU-accelerated inference to address current timeout rates on CPU hardware.

Membership inference attacks (MIAs) have been extensively studied in large language models (LLMs) and vision-language models (VLMs), yet their implications for vision-language-action (VLA) models remain largely unexplored. VLA models differ from standard LLMs and VLMs in several important ways: they are often fine-tuned for many epochs on relatively small embodied datasets, operate over constrained and structured action spaces, and expose action outputs that can be observed as executable behaviors and temporally correlated trajectories. These characteristics suggest a distinct and potentially more informative attack surface for membership inference. In this work, we present the first systematic study of MIAs against VLA systems. We formalize two membership inference settings for VLA models: sample-level inference over individual transition samples and trajectory-level inference over complete embodied demonstrations. We further develop a suite of attack methods under multiple access regimes, including strict black-box access. Our attacks exploit both classic MIA signals, such as token likelihood, and VLA-specific signals, such as observable action errors and temporal motion patterns. Across multiple VLA benchmarks and representative VLA models, these attacks achieve strong inference performance, showing that VLA models are highly vulnerable to membership inference. Notably, black-box attacks based only on generated actions achieve strong performance, highlighting a practical privacy risk for deployed embodied AI systems. Our findings reveal a previously underexplored privacy risk in robotic and embodied AI, and underscore the need for dedicated privacy evaluation and defenses for VLA models.

Energy-based models (EBMs) implement inference as gradient descent on a learned Lyapunov function, yielding interpretable, structure-preserving alternatives to black-box neural ODEs and aligning naturally with physical AI. Yet their use in system identification remains limited, and existing architectures lack formal stability guarantees that globally preclude unstable modes. We address this gap by introducing an EBM framework for system identification with stable, dissipative, absorbing invariant dynamics. Unlike classical global Lyapunov stability, absorbing invariance expands the class of stability-preserving architectures, enabling more flexible and expressive EBMs. We extend EBM theory to nonsmooth activations by establishing negative energy dissipation via Clarke derivatives and deriving new conditions for radial unboundedness, exposing a stability-expressivity tradeoff in standard EBMs. To overcome this, we introduce a hybrid architecture with a dynamical visible layer and static hidden layers, prove absorbing invariance under mild assumptions, and show that these guarantees extend to port-Hamiltonian EBMs. Experiments on metric-deformed multi-well and ring systems validate the approach, showcasing how our hybrid EBM architecture combines expressivity with sound and provable safety guarantees by design.

Large language models (LLMs) are increasingly used as simulated participants in social science experiments, but their behavior is often unstable and highly sensitive to design choices. Prior evaluations frequently conflate base-model capabilities with experimental instantiation, obscuring whether outcomes reflect the model itself or the agent setup. We instead frame participant simulation as an agent-design problem over full experimental protocols, where an agent is defined by a base model and a specification (e.g., participant attributes) that encodes behavioral assumptions. We introduce HUMANSTUDY-BENCH, a benchmark and execution engine that orchestrates LLM-based agents to reconstruct published human-subject experiments via a Filter--Extract--Execute--Evaluate pipeline, replaying trial sequences and running the original analysis pipeline in a shared runtime that preserves the original statistical procedures end to end. To evaluate fidelity at the level of scientific inference, we propose new metrics to quantify how much human and agent behaviors agree. We instantiate 12 foundational studies as an initial suite in this dynamic benchmark, spanning individual cognition, strategic interaction, and social psychology, and covering more than 6,000 trials with human samples ranging from tens to over 2,100 participants.

Effective long-term memory in conversational AI requires synthesizing information across multiple sessions. However, current systems place excessive reasoning burden on response generation, making performance significantly dependent on model sizes. We introduce PREMem (Pre-storage Reasoning for Episodic Memory), a novel approach that shifts complex reasoning processes from inference to memory construction. PREMem extracts fine-grained memory fragments categorized into factual, experiential, and subjective information; it then establishes explicit relationships between memory items across sessions, capturing evolution patterns like extensions, transformations, and implications. By performing this reasoning during pre-storage rather than when generating a response, PREMem creates enriched representations while reducing computational demands during interactions. Experiments show significant performance improvements across all model sizes, with smaller models achieving results comparable to much larger baselines while maintaining effectiveness even with constrained token budgets. Code and dataset are available at https://github.com/sangyeop-kim/PREMem.

In computational cognitive modeling, capturing the full spectrum of human judgment and decision-making processes, beyond just optimal behaviors, is a significant challenge. This study explores whether Large Language Models (LLMs) can emulate the breadth of human reasoning by predicting both intuitive, fast System 1 and deliberate, slow System 2 processes. We investigate the potential of AI to mimic diverse reasoning behaviors across a human population, addressing what we call the "full reasoning spectrum problem". We designed reasoning tasks using a novel generalization of the Natural Language Inference (NLI) format to evaluate LLMs' ability to replicate human reasoning. The questions were crafted to elicit both System 1 and System 2 responses. Human responses were collected through crowd-sourcing and the entire distribution was modeled, rather than just the majority of the answers. We used personality-based prompting inspired by the Big Five personality model to elicit AI responses reflecting specific personality traits, capturing the diversity of human reasoning, and exploring how personality traits influence LLM outputs. Combined with genetic algorithms to optimize the weighting of these prompts, this method was tested alongside traditional machine learning models. The results show that LLMs can mimic human response distributions, with open-source models like Llama and Mistral outperforming proprietary GPT models. Personality-based prompting, especially when optimized with genetic algorithms, significantly enhanced LLMs' ability to predict human response distributions, suggesting that capturing suboptimal, naturalistic reasoning may require modeling techniques incorporating diverse reasoning styles and psychological profiles. The study concludes that personality-based prompting combined with genetic algorithms is promising for enhancing AI's 'human-ness' in reasoning.

Generative artificial intelligence (AI) technology is revolutionizing the computing industry. Not only its applications have broadened to various sectors but also poses new system design and optimization opportunities. The technology is capable of understanding and responding in multiple modalities. However, the advanced capability currently comes with significant system resource demands. To sustainably scale generative AI capabilities to billions of users in the world, inference must be fast and efficient. This paper pinpoints key system design and optimization opportunities by characterizing a family of emerging multi-modal generation models on real systems. Auto-regressive token generation is a critical latency performance bottleneck, typically dominated by GPU idle time. In addition to memory-intensive attention across the generative AI models, linear operations constitute significant inference latency due to the feed forward networks in Transformer-based models. We demonstrate that state-of-the-art optimization levers, spanning from applications to system software and hardware, set a 3.88x better baseline.

Large Language Models(LLMs) have had a profound impact on AI applications, particularly in the domains of long-text comprehension and generation. KV Cache technology is one of the most widely used techniques in the industry. It ensures efficient sequence generation by caching previously computed KV states. However, it also introduces significant memory overhead. We discovered that KV Cache is not necessary and proposed a novel KCache technique to alleviate the memory bottleneck issue during the LLMs inference process. KCache can be used directly for inference without any training process, Our evaluations show that KCache improves the throughput of popular LLMs by 40% with the baseline, while keeping accuracy.

We propose Embodied AI as the next fundamental step in the pursuit of Artificial General Intelligence, juxtaposing it against current AI advancements, particularly Large Language Models. We traverse the evolution of the embodiment concept across diverse fields - philosophy, psychology, neuroscience, and robotics - to highlight how EAI distinguishes itself from the classical paradigm of static learning. By broadening the scope of Embodied AI, we introduce a theoretical framework based on cognitive architectures, emphasizing perception, action, memory, and learning as essential components of an embodied agent. This framework is aligned with Friston's active inference principle, offering a comprehensive approach to EAI development. Despite the progress made in the field of AI, substantial challenges, such as the formulation of a novel AI learning theory and the innovation of advanced hardware, persist. Our discussion lays down a foundational guideline for future Embodied AI research. Highlighting the importance of creating Embodied AI agents capable of seamless communication, collaboration, and coexistence with humans and other intelligent entities within real-world environments, we aim to steer the AI community towards addressing the multifaceted challenges and seizing the opportunities that lie ahead in the quest for AGI.

Large Language Models (LLMs) for Generative AI have achieved remarkable progress, evolving into sophisticated and versatile tools widely adopted across various domains and applications. However, the substantial memory overhead caused by their vast number of parameters, combined with the high computational demands of the attention mechanism, poses significant challenges in achieving low latency and high throughput for LLM inference services. Recent advancements, driven by groundbreaking research, have significantly accelerated progress in this field. This paper provides a comprehensive survey of these methods, covering fundamental instance-level approaches, in-depth cluster-level strategies, emerging scenario directions, and other miscellaneous but important areas. At the instance level, we review model placement, request scheduling, decoding length prediction, storage management, and the disaggregation paradigm. At the cluster level, we explore GPU cluster deployment, multi-instance load balancing, and cloud service solutions. For emerging scenarios, we organize the discussion around specific tasks, modules, and auxiliary methods. To ensure a holistic overview, we also highlight several niche yet critical areas. Finally, we outline potential research directions to further advance the field of LLM inference serving.

Recent advances in multimodal web agents often rely on increased inference-time computation, including rollout search, verifier passes, offline skill discovery, and specialist model stacks. This raises a central question: can a web agent become more efficient as it accumulates experience, rather than more expensive? We first analyze trajectories from VisualWebArena and identify three recurring sources of inefficiency: repeat-action loops, hidden discovery costs, and low prompt-cache reuse. We then introduce PANDO, a single-rollout online skill-distillation framework that maintains a structured Skill Library and combines progress reflection, confidence-based skill demotion, hierarchical routing, visual compression, and cache-aware prompting. On the full set of 910 VisualWebArena tasks, PANDO achieves a 58.3% success rate, outperforming SGV (54.0%) and our WALT reproduction (45.2%), while using 58% fewer tokens than SGV and 61% fewer tokens than WALT, without any pre-evaluation discovery budget. A 300-task ablation further shows that rules and routines provide most of the success gains, while routing, compression, and cache-aware prompting convert the larger skill library into lower marginal token cost. Finally, we introduce three trajectory-level efficiency metrics -- Action Repetition Rate, Step Overhead Ratio, and Prompt Cache Utilization -- to make efficiency visible beyond terminal success.

Controllable music generation methods are critical for human-centered AI-based music creation, but are currently limited by speed, quality, and control design trade-offs. Diffusion Inference-Time T-optimization (DITTO), in particular, offers state-of-the-art results, but is over 10x slower than real-time, limiting practical use. We propose Distilled Diffusion Inference-Time T -Optimization (or DITTO-2), a new method to speed up inference-time optimization-based control and unlock faster-than-real-time generation for a wide-variety of applications such as music inpainting, outpainting, intensity, melody, and musical structure control. Our method works by (1) distilling a pre-trained diffusion model for fast sampling via an efficient, modified consistency or consistency trajectory distillation process (2) performing inference-time optimization using our distilled model with one-step sampling as an efficient surrogate optimization task and (3) running a final multi-step sampling generation (decoding) using our estimated noise latents for best-quality, fast, controllable generation. Through thorough evaluation, we find our method not only speeds up generation over 10-20x, but simultaneously improves control adherence and generation quality all at once. Furthermore, we apply our approach to a new application of maximizing text adherence (CLAP score) and show we can convert an unconditional diffusion model without text inputs into a model that yields state-of-the-art text control. Sound examples can be found at https://ditto-music.github.io/ditto2/.

Tool-augmented large language models (LLMs) are emerging as deep research agents, systems that decompose complex queries, retrieve external evidence, and synthesize grounded responses. Yet current agents remain limited by shallow retrieval, weak alignment metrics, and brittle tool-use behavior. We introduce PokeeResearch-7B, a 7B-parameter deep research agent built under a unified reinforcement learning framework for robustness, alignment, and scalability. PokeeResearch-7B is trained by an annotation-free Reinforcement Learning from AI Feedback (RLAIF) framework to optimize policies using LLM-based reward signals that capture factual accuracy, citation faithfulness, and instruction adherence. A chain-of-thought-driven multi-call reasoning scaffold further enhances robustness through self-verification and adaptive recovery from tool failures. Among 10 popular deep research benchmarks, PokeeResearch-7B achieves state-of-the-art performance among 7B-scale deep research agents. This highlights that careful reinforcement learning and reasoning design can produce efficient, resilient, and research-grade AI agents. The model and inference code is open-sourced under MIT license at https://github.com/Pokee-AI/PokeeResearchOSS.

As the development of large-scale Generative AI models evolve beyond text (1D) generation to include image (2D) and video (3D) generation, processing spatial and temporal information presents unique challenges to quality, performance, and efficiency. We present the first work towards understanding this new system design space for multi-modal text-to-image (TTI) and text-to-video (TTV) generation models. Current model architecture designs are bifurcated into 2 categories: Diffusion- and Transformer-based models. Our systematic performance characterization on a suite of eight representative TTI/TTV models shows that after state-of-the-art optimization techniques such as Flash Attention are applied, Convolution accounts for up to 44% of execution time for Diffusion-based TTI models, while Linear layers consume up to 49% of execution time for Transformer-based models. We additionally observe that Diffusion-based TTI models resemble the Prefill stage of LLM inference, and benefit from 1.1-2.5x greater speedup from Flash Attention than Transformer-based TTI models that resemble the Decode phase. Since optimizations designed for LLMs do not map directly onto TTI/TTV models, we must conduct a thorough characterization of these workloads to gain insights for new optimization opportunities. In doing so, we define sequence length in the context of TTI/TTV models and observe sequence length can vary up to 4x in Diffusion model inference. We additionally observe temporal aspects of TTV workloads pose unique system bottlenecks, with Temporal Attention accounting for over 60% of total Attention time. Overall, our in-depth system performance characterization is a critical first step towards designing efficient and deployable systems for emerging TTI/TTV workloads.

Making deep learning recommendation model (DLRM) training and inference fast and efficient is important. However, this presents three key system challenges - model architecture diversity, kernel primitive diversity, and hardware generation and architecture heterogeneity. This paper presents KernelEvolve-an agentic kernel coding framework-to tackle heterogeneity at-scale for DLRM. KernelEvolve is designed to take kernel specifications as input and automate the process of kernel generation and optimization for recommendation model across heterogeneous hardware architectures. KernelEvolve does so by operating at multiple programming abstractions, from Triton and CuTe DSL to low-level hardware agnostic languages, spanning the full hardware-software optimization stack. The kernel optimization process is described as graph-based search with selection policy, universal operator, fitness function, and termination rule, dynamically adapts to runtime execution context through retrieval-augmented prompt synthesis. We designed, implemented, and deployed KernelEvolve to optimize a wide variety of production recommendation models across generations of NVIDIA and AMD GPUs, as well as Meta's AI accelerators. We validate KernelEvolve on the publicly-available KernelBench suite, achieving 100% pass rate on all 250 problems across three difficulty levels, and 160 PyTorch ATen operators across three heterogeneous hardware platforms, demonstrating 100% correctness. KernelEvolve reduces development time from weeks to hours and achieves substantial performance improvements over PyTorch baselines across diverse production use cases and for heterogeneous AI systems at-scale. Beyond performance efficiency improvements, KernelEvolve significantly mitigates the programmability barrier for new AI hardware by enabling automated kernel generation for in-house developed AI hardware.

Evaluating generative AI models is increasingly resource-intensive due to slow inference, expensive raters, and a rapidly growing landscape of models and benchmarks. We propose ProEval, a proactive evaluation framework that leverages transfer learning to efficiently estimate performance and identify failure cases. ProEval employs pre-trained Gaussian Processes (GPs) as surrogates for the performance score function, mapping model inputs to metrics such as the severity of errors or safety violations. By framing performance estimation as Bayesian quadrature (BQ) and failure discovery as superlevel set sampling, we develop uncertainty-aware decision strategies that actively select or synthesize highly informative inputs for testing. Theoretically, we prove that our pre-trained GP-based BQ estimator is unbiased and bounded. Empirically, extensive experiments on reasoning, safety alignment, and classification benchmarks demonstrate that ProEval is significantly more efficient than competitive baselines. It requires 8-65x fewer samples to achieve estimates within 1% of the ground truth, while simultaneously revealing more diverse failure cases under a stricter evaluation budget.

Optimizing inference for long-context Large Language Models (LLMs) is increasingly important due to the quadratic compute and linear memory complexity of Transformers. Existing approximation methods, such as key-value (KV) cache dropping, sparse attention, and prompt compression, typically rely on rough predictions of token or KV pair importance. We propose a novel framework for approximate LLM inference that leverages small draft models to more accurately predict the importance of tokens and KV pairs. Specifically, we introduce two instantiations of our proposed framework: (i) SpecKV, which leverages a draft output to accurately assess the importance of each KV pair for more effective KV cache dropping, and (ii) SpecPC, which uses the draft model's attention activations to identify and discard unimportant prompt tokens. To the best of our knowledge, this is the first work to use draft models for approximate LLM inference acceleration, extending their utility beyond traditional lossless speculative decoding. We motivate our methods with theoretical and empirical analyses, and show a strong correlation between the attention patterns of draft and target models. Extensive experiments on long-context benchmarks show that our methods consistently achieve higher accuracy than existing baselines, while preserving the same improvements in memory usage, latency, and throughput. Our code is available at https://github.com/furiosa-ai/draft-based-approx-llm.

Large language models run every token through every layer, regardless of difficulty. We present TIDE, a post-training system that attaches tiny learned routers at periodic checkpoint layers and, at inference time, selects the earliest layer whose hidden state has converged for each token. TIDE requires no model retraining, works with any HuggingFace causal LM, auto-detects GPU architecture, and supports float32, float16, and bfloat16 through fused CUDA kernels. On an NVIDIA A100 with DeepSeek R1 Distill 8B, TIDE achieves 100% prefill exit rate (5% of tokens exit at layer 11, the remaining at layer 31), reduces prefill latency by 7.2%, and increases single-batch throughput by 6.6%. During autoregressive decoding, 98-99% of tokens exit early while the model correctly solves a multi-step math problem with 95 unique output tokens. On Qwen3 8B (36 layers), throughput improves by 8.1% at batch size 8. Calibration on 2,000 WikiText samples takes under 3 minutes and produces a ~4 MB router checkpoint. The system comprises 1,308 lines of Python and 1,081 lines of CUDA/C++ with 74 passing tests. Code: https://github.com/RightNow-AI/TIDE

Can an AI autonomously design mechanisms for computer systems on par with the creativity and reasoning of human experts? We present Glia, an AI architecture for networked systems design that uses large language models (LLMs) in a human-inspired, multi-agent workflow. Each agent specializes in reasoning, experimentation, and analysis, collaborating through an evaluation framework that grounds abstract reasoning in empirical feedback. Unlike prior ML-for-systems methods that optimize black-box policies, Glia generates interpretable designs and exposes its reasoning process. When applied to a distributed GPU cluster for LLM inference, it produces new algorithms for request routing, scheduling, and auto-scaling that perform at human-expert levels in significantly less time, while yielding novel insights into workload behavior. Our results suggest that by combining reasoning LLMs with structured experimentation, an AI can produce creative and understandable designs for complex systems problems.

Urban causal research is essential for understanding the complex dynamics of cities and informing evidence-based policies. However, it is challenged by the inefficiency and bias of hypothesis generation, barriers to multimodal data complexity, and the methodological fragility of causal experimentation. Recent advances in large language models (LLMs) present an opportunity to rethink how urban causal analysis is conducted. This Perspective examines current urban causal research by analyzing taxonomies that categorize research topics, data sources, and methodological approaches to identify structural gaps. We then introduce an LLM-driven conceptual framework, AutoUrbanCI, composed of four distinct modular agents responsible for hypothesis generation, data engineering, experiment design and execution, and results interpretation with policy recommendations. We propose evaluation criteria for rigor and transparency and reflect on implications for human-AI collaboration, equity, and accountability. We call for a new research agenda that embraces AI-augmented workflows not as replacements for human expertise but as tools to broaden participation, improve reproducibility, and unlock more inclusive forms of urban causal reasoning.

Agentic systems built on large language models (LLMs) extend beyond text generation to autonomously retrieve information and invoke tools. This runtime execution model shifts the attack surface from build-time artifacts to inference-time dependencies, exposing agents to manipulation through untrusted data and probabilistic capability resolution. While prior work has focused on model-level vulnerabilities, security risks emerging from cyclic and interdependent runtime behavior remain fragmented. We systematize these risks within a unified runtime framework, categorizing threats into data supply chain attacks (transient context injection and persistent memory poisoning) and tool supply chain attacks (discovery, implementation, and invocation). We further identify the Viral Agent Loop, in which agents act as vectors for self-propagating generative worms without exploiting code-level flaws. Finally, we advocate a Zero-Trust Runtime Architecture that treats context as untrusted control flow and constrains tool execution through cryptographic provenance rather than semantic inference.

With the rapid advancement of artificial intelligence technologies such as ChatGPT, AI agents, and video generation, contemporary mobile systems have begun integrating these AI capabilities on local devices to enhance privacy and reduce response latency. To meet the computational demands of AI tasks, current mobile SoCs are equipped with diverse AI accelerators, including GPUs and Neural Processing Units (NPUs). However, there has not been a comprehensive characterization of these heterogeneous processors, and existing designs typically only leverage a single AI accelerator for LLM inference, leading to suboptimal use of computational resources and memory bandwidth. In this paper, we first summarize key performance characteristics of heterogeneous processors, SoC memory bandwidth, etc. Drawing on these observations, we propose different heterogeneous parallel mechanisms to fully exploit both GPU and NPU computational power and memory bandwidth. We further design a fast synchronization mechanism between heterogeneous processors that leverages the unified memory architecture. By employing these techniques, we present HeteroInfer, the fastest LLM inference engine in mobile devices which supports GPU-NPU heterogeneous execution. Evaluation shows that HeteroInfer delivers a 1.34x to 6.02x end-to-end speedup over state-of-the-art GPU-only and NPU-only LLM engines, while maintaining negligible interference with other applications.

The energy consumption of large-scale ML models is dominated by data movement - shuffling billions of parameters across memory hierarchies and data centers. Effective sparsification to prune redundant parameters is still challenging: existing methods incur significant accuracy degradation, performance overhead, or both. We introduce (Bl)ock (a)nd (S)parse (T)ransformers (BLaST), a general, robust, and reliable sparsification method applicable to linear layers in all settings. Our method iteratively sparsifies weight matrices into a block sparsity pattern suitable for efficient sparse matrix-matrix (SpMM) multiplication. BLaST achieves up to 95% sparsity in MLP weights with negligible accuracy loss. Our fused, highly optimized Sparse MLP kernel delivers up to 16.7x speedup over dense MLPs across 9 architectures and 8 datasets, resulting in up to 1.6x inference speedup, 1.11x pretraining speedup and up to 3.12x inference memory usage reduction. BLaST enables the next generation of large-scale AI systems by reducing energy use, memory footprint, and latency.

Transformer-based Large Language Models (LLMs) have significantly advanced AI capabilities but pose considerable challenges for deployment on edge devices due to high computational demands, memory bandwidth constraints, and energy consumption. This paper addresses these challenges by presenting an efficient framework for deploying the Qwen2.5-0.5B model on the Xilinx Kria KV260 edge platform, a heterogeneous system integrating an ARM Cortex-A53 CPU with reconfigurable FPGA logic. Leveraging Activation-aware Weight Quantization (AWQ) with FPGA-accelerated execution pipelines, the proposed approach enhances both model compression rate and system throughput. Additionally, we propose a hybrid execution strategy that intelligently offloads compute-intensive operations to the FPGA while utilizing the CPU for lighter tasks, effectively balancing the computational workload and maximizing overall performance. Our framework achieves a model compression rate of 55.08% compared to the original model and produces output at a rate of 5.1 tokens per second, outperforming the baseline performance of 2.8 tokens per second.

Energy efficiency of Convolutional Neural Networks (CNNs) has become an important area of research, with various strategies being developed to minimize the power consumption of these models. Previous efforts, including techniques like model pruning, quantization, and hardware optimization, have made significant strides in this direction. However, there remains a need for more effective on device AI solutions that balance energy efficiency with model performance. In this paper, we propose a novel approach to reduce the energy requirements of inference of CNNs. Our methodology employs two small Complementary CNNs that collaborate with each other by covering each other's "weaknesses" in predictions. If the confidence for a prediction of the first CNN is considered low, the second CNN is invoked with the aim of producing a higher confidence prediction. This dual-CNN setup significantly reduces energy consumption compared to using a single large deep CNN. Additionally, we propose a memory component that retains previous classifications for identical inputs, bypassing the need to re-invoke the CNNs for the same input, further saving energy. Our experiments on a Jetson Nano computer demonstrate an energy reduction of up to 85.8% achieved on modified datasets where each sample was duplicated once. These findings indicate that leveraging a complementary CNN pair along with a memory component effectively reduces inference energy while maintaining high accuracy.

Isambard-AI is a new, leadership-class supercomputer, designed to support AI-related research. Based on the HPE Cray EX4000 system, and housed in a new, energy efficient Modular Data Centre in Bristol, UK, Isambard-AI employs 5,448 NVIDIA Grace-Hopper GPUs to deliver over 21 ExaFLOP/s of 8-bit floating point performance for LLM training, and over 250 PetaFLOP/s of 64-bit performance, for under 5MW. Isambard-AI integrates two, all-flash storage systems: a 20 PiByte Cray ClusterStor and a 3.5 PiByte VAST solution. Combined these give Isambard-AI flexibility for training, inference and secure data accesses and sharing. But it is the software stack where Isambard-AI will be most different from traditional HPC systems. Isambard-AI is designed to support users who may have been using GPUs in the cloud, and so access will more typically be via Jupyter notebooks, MLOps, or other web-based, interactive interfaces, rather than the approach used on traditional supercomputers of sshing into a system before submitting jobs to a batch scheduler. Its stack is designed to be quickly and regularly upgraded to keep pace with the rapid evolution of AI software, with full support for containers. Phase 1 of Isambard-AI is due online in May/June 2024, with the full system expected in production by the end of the year.

An increasing number of studies use gender information to understand phenomena such as gender bias, inequity in access and participation, or the impact of the Covid pandemic response. Unfortunately, most datasets do not include self-reported gender information, making it necessary for researchers to infer gender from other information, such as names or names and country information. An important limitation of these tools is that they fail to appropriately capture the fact that gender exists on a non-binary scale, however, it remains important to evaluate and compare how well these tools perform in a variety of contexts. In this paper, we compare the performance of a generative Artificial Intelligence (AI) tool ChatGPT with three commercially available list-based and machine learning-based gender inference tools (Namsor, Gender-API, and genderize.io) on a unique dataset. Specifically, we use a large Olympic athlete dataset and report how variations in the input (e.g., first name and first and last name, with and without country information) impact the accuracy of their predictions. We report results for the full set, as well as for the subsets: medal versus non-medal winners, athletes from the largest English-speaking countries, and athletes from East Asia. On these sets, we find that Namsor is the best traditional commercially available tool. However, ChatGPT performs at least as well as Namsor and often outperforms it, especially for the female sample when country and/or last name information is available. All tools perform better on medalists versus non-medalists and on names from English-speaking countries. Although not designed for this purpose, ChatGPT may be a cost-effective tool for gender prediction. In the future, it might even be possible for ChatGPT or other large scale language models to better identify self-reported gender rather than report gender on a binary scale.

Web applications are increasingly becoming the primary platform for AI service delivery, making in-browser deep learning (DL) inference more prominent. However, current in-browser inference systems fail to effectively utilize advanced web programming techniques and customize kernels for various client devices, leading to suboptimal performance. To address the issues, this paper presents the first in-browser inference system, nn-JIT.web, which enables just-in-time (JIT) auto-generation of optimized kernels for both CPUs and GPUs during inference. The system achieves this by using two novel web programming techniques that can significantly reduce kernel generation time, compared to other tensor compilers such as TVM, while maintaining or even improving performance. The first technique, Tensor-Web Compiling Co-Design, lowers compiling costs by unifying tensor and web compiling and eliminating redundant and ineffective compiling passes. The second technique, Web-Specific Lite Kernel Optimization Space Design, reduces kernel tuning costs by focusing on web programming requirements and efficient hardware resource utilization, limiting the optimization space to only dozens. nn-JIT.web is evaluated for modern transformer models on a range of client devices, including the mainstream CPUs and GPUs from ARM, Intel, AMD and Nvidia. Results show that nn-JIT.web can achieve up to 8.2x faster within 30 seconds compared to the baselines across various models.

Large foundation models have demonstrated a great ability to achieve general human-level intelligence far beyond traditional approaches. As the technique keeps attracting attention from the AI community, more and more large foundation models have become publically available. However, most of those models exhibit a major deficiency in specialized-task applications, where the step of finetuning is still required for obtaining satisfactory performance. As the number of available models and specialized tasks keeps growing, the job of general finetuning becomes highly nontrivial. In this paper, we take the first step to address this issue. We introduce an extensible and lightweight toolkit, LMFlow, which aims to simplify the finetuning and inference of general large foundation models. LMFlow offers a complete finetuning workflow for a large foundation model to support personalized training with limited computing resources. Furthermore, it supports continuous pretraining, instruction tuning, parameter-efficient finetuning, alignment tuning, and large model inference, along with carefully designed and extensible APIs. This toolkit has been thoroughly tested and is available at https://github.com/OptimalScale/LMFlow.