🔄 In a Training Loop

Abdullah Mujeeb Khawaja

amkkk

11 2 36

https://navigatingnoise.substack.com/

AI & ML interests

Tough question as I love all areas of DL, but from all of it, embeddings are my favourite topic. Currently obsessed with what can we do around LLMs. MLE @ DPD, prev DS in Recs Systems @ Alibaba Group

Recent Activity

updated a model 1 day ago

amkkk/Trace-Inverter-4B-NoBubble

liked a model 1 day ago

arcee-ai/AFM-4.5B-Base-KDA-NoPE

liked a model 2 days ago

arcee-ai/AFM-4.5B-Base

View all activity

Organizations

amkkk 's Spaces 283

Reproduction: A Fine-Grained Understanding of Uniform Convergence for Halfspaces

🎯

Explore code logs, traces and workspace in an interactive logbook

Reproduction: Zero-Flow Encoders

🔬

Explore project logs, traces, and workspace online

Reproduction: How to Correctly Report LLM-as-a-Judge Evaluations

📊

Explore and manage experiment logs with code, traces, and workspace

Reproduction: Unifying and Optimizing Data Values for Selection via Sequential Decision-Making

🏅

View and manage experiment logs, traces, and workspace

Reproduction: Privacy Amplification in Differentially Private Zeroth-Order Optimization with Hidden States

🔒

Explore research logbooks with code, traces, and workspace

Reproduction: Improved Dimension Dependence for Bandit Convex Optimization with Gradient Variations

🎯

Explore code logs, traces, and workspace with a collaborative logbook

Reproduction: Online Robust Reinforcement Learning with General Function Approximation

🛡

Explore reinforcement learning experiment logs online

Reproduction: Repro - Robust Linear Dueling Bandits with Post-serving Context under Unknown Delays and Adversarial Corruptions

🎯

Explore code, traces, and workspace in an interactive logbook

Reproduction: Recursive Binding on a Budget: Subspace Carving in Order-p Tensor Memories

🎯

Explore project logs, traces, and workspace in a web UI

Reproduction: Adaptive Multi-Round Allocation with Stochastic Arrivals

🎯

Explore research logbook with code, traces, and workspace

Reproduction: New Bounds for Kernel Sums via Fast Spherical Embeddings

🎯

View code logs, traces and workspace with collaborative support

Reproduction: INDUCTION: Finite-Structure Concept Synthesis in First-Order Logic

🎯

Explore project logs, traces, and workspace with AI collaboration

Reproduction: Towards Optimal Robustness in Learning-Augmented Paging

🎯

Explore project logs, traces, and workspace in an interactive view

Reproduction: Infinite-Dimensional Generative Diffusions via Doob's h-Transform

🎯

Explore research logbooks with code, traces, and workspace

Reproduction: Efficient Inference for Noisy LLM-as-a-Judge Evaluation

🎯

Explore experiment logs, traces, and workspace files

Reproduction: On the Convergence Rate of LoRA Gradient Descent

📉

Explore experiment logs, code, and traces in an interactive UI

Reproduction: Improved Distribution Estimation in ℓ∞

📊

Explore code logs, traces, and workspace in an interactive UI

Reproduction: Optimal Rates for Feasible Payoff Set Estimation in Games

🎯

Explore code, traces, and workspace in an interactive logbook

Reproduction: Nash Equilibria in Games with Playerwise Concave Coupling Constraints

🎯

Explore project logs, traces, and workspace in an interactive UI

Reproduction: Explanations are a Means to an End: Decision Theoretic Explanation Evaluation

🎯

Explore code logs, execution traces, and workspace

Reproduction: Adaptive Estimation and Inference in Semi-parametric Heterogeneous Clustered Multitask Learning via Neyman Orthogonality

🧪

Explore code, traces, and workspace in an interactive logbook

Reproduction: Tight Margin-Based Generalization Bounds for Voting Classifiers over Finite Hypothesis Sets

🎯

Explore and manage code logs, traces, and workspace

Reproduction: Language Generation with Replay: A Learning-Theoretic View of Model Collapse

🔁

Explore experiment logs and collaborate with an AI agent

Reproduction: FIBER: A Differentially Private Optimizer with Filter-Aware Innovation Bias Correction

🎯

Explore project code, traces, and workspace in an interactive logbook

Abdullah Mujeeb Khawaja

AI & ML interests

Recent Activity

Organizations

amkkk 's Spaces 283 Sort: Recently updated

Reproduction: A Fine-Grained Understanding of Uniform Convergence for Halfspaces

Reproduction: Zero-Flow Encoders

Reproduction: How to Correctly Report LLM-as-a-Judge Evaluations

Reproduction: Unifying and Optimizing Data Values for Selection via Sequential Decision-Making

Reproduction: Privacy Amplification in Differentially Private Zeroth-Order Optimization with Hidden States

Reproduction: Improved Dimension Dependence for Bandit Convex Optimization with Gradient Variations

Reproduction: Online Robust Reinforcement Learning with General Function Approximation

Reproduction: Repro - Robust Linear Dueling Bandits with Post-serving Context under Unknown Delays and Adversarial Corruptions

Reproduction: Recursive Binding on a Budget: Subspace Carving in Order-p Tensor Memories

Reproduction: Adaptive Multi-Round Allocation with Stochastic Arrivals

Reproduction: New Bounds for Kernel Sums via Fast Spherical Embeddings

Reproduction: INDUCTION: Finite-Structure Concept Synthesis in First-Order Logic

Reproduction: Towards Optimal Robustness in Learning-Augmented Paging

Reproduction: Infinite-Dimensional Generative Diffusions via Doob's h-Transform

Reproduction: Efficient Inference for Noisy LLM-as-a-Judge Evaluation

Reproduction: On the Convergence Rate of LoRA Gradient Descent

Reproduction: Improved Distribution Estimation in ℓ∞

Reproduction: Optimal Rates for Feasible Payoff Set Estimation in Games

Reproduction: Nash Equilibria in Games with Playerwise Concave Coupling Constraints

Reproduction: Explanations are a Means to an End: Decision Theoretic Explanation Evaluation

Reproduction: Adaptive Estimation and Inference in Semi-parametric Heterogeneous Clustered Multitask Learning via Neyman Orthogonality

Reproduction: Tight Margin-Based Generalization Bounds for Voting Classifiers over Finite Hypothesis Sets

Reproduction: Language Generation with Replay: A Learning-Theoretic View of Model Collapse

Reproduction: FIBER: A Differentially Private Optimizer with Filter-Aware Innovation Bias Correction

amkkk 's Spaces 283