google/gemma-4-26B-A4B-it
Image-Text-to-Text • 27B • Updated • 836k • • 527
Aleksander Wlodarczyk
alexw404
·
AI & ML interests
None yet
Recent Activity
liked a model 1 day ago
google/gemma-4-26B-A4B-it liked a model 1 day ago
google/gemma-4-26B-A4B liked a model 2 days ago
google/gemma-4-31B-itOrganizations
None yet
liked 2
models 1 day ago
liked a
model 2 days ago
liked 2
models 4 days ago
liked 6
models 6 days ago
Jackrong/Qwopus3.5-27B-v3-GGUF
Image-Text-to-Text • 27B • Updated • 64.3k • 227
chromadb/context-1
Text Generation • 21B • Updated • 4.05k • 377
mradermacher/Holo3-35B-A3B-GGUF
35B • Updated • 5.17k • 10
Hcompany/Holo3-35B-A3B
Image-Text-to-Text • 35B • Updated • 1.83k • 253
mradermacher/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-GGUF
27B • Updated • 32.4k • 25
Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled
Image-Text-to-Text • 28B • Updated • 561k • 2.48k
reactedto reaperdoesntknow's post with 👍 6 days ago
Post
1976
Your Loss Function Has Singularities. Classical Calculus Can't See Them.
Introducing Discrepancy Calculus (DISC) — treating training singularities as structure, not noise.
Loss plateaus, mode collapse, catastrophic forgetting, distilled models that know things the teacher never taught — we engineer around these. But what if those singularities are the actual structure of the learning problem?
The core insight: Every BV function decomposes into smooth (what classical calculus handles), jump (capability emergence, loss plateaus breaking), and Cantor (ghost imprinting — knowledge transferring through weight-space topology, not gradient signal). Classical analysis sees only the first. DISC sees all three.
The paper proves this isn't alternative notation — it's strictly larger. The Meta-Discrepancy Theorem: where singularities exist, the classical FTC/MVT/chain-rule package is provably impossible.
What it explains:
TopologicalQwen exhibited literary reasoning from physics-only data — the Cantor part explains how. DualMind's Explore→Examine→Response loop operationalizes DISC as inference dynamics. 50 models, 35K+ downloads, all built on this framework.
Paper: Discrepancy Calculus: Foundations and Core Theory (DOI: 10.57967/hf/8194) — 8 axioms, proofs, computational recipes.
Series: Structure Over Scale (DOI: 10.57967/hf/8165) → Three Teachers to Dual Cognition (DOI: 10.57967/hf/8184) → DISC Foundations
— Roy S. Colca Jr., Convergent Intelligence LLC: Research Division
Introducing Discrepancy Calculus (DISC) — treating training singularities as structure, not noise.
Loss plateaus, mode collapse, catastrophic forgetting, distilled models that know things the teacher never taught — we engineer around these. But what if those singularities are the actual structure of the learning problem?
The core insight: Every BV function decomposes into smooth (what classical calculus handles), jump (capability emergence, loss plateaus breaking), and Cantor (ghost imprinting — knowledge transferring through weight-space topology, not gradient signal). Classical analysis sees only the first. DISC sees all three.
The paper proves this isn't alternative notation — it's strictly larger. The Meta-Discrepancy Theorem: where singularities exist, the classical FTC/MVT/chain-rule package is provably impossible.
What it explains:
TopologicalQwen exhibited literary reasoning from physics-only data — the Cantor part explains how. DualMind's Explore→Examine→Response loop operationalizes DISC as inference dynamics. 50 models, 35K+ downloads, all built on this framework.
Paper: Discrepancy Calculus: Foundations and Core Theory (DOI: 10.57967/hf/8194) — 8 axioms, proofs, computational recipes.
Series: Structure Over Scale (DOI: 10.57967/hf/8165) → Three Teachers to Dual Cognition (DOI: 10.57967/hf/8184) → DISC Foundations
— Roy S. Colca Jr., Convergent Intelligence LLC: Research Division
liked 6
models 6 days ago
ibm-granite/granite-4.0-3b-vision
Image-Text-to-Text • 4B • Updated • 8.01k • 86
mistralai/Voxtral-4B-TTS-2603
Text-to-Speech • Updated • 6.19k • 695
microsoft/harrier-oss-v1-27b
Feature Extraction • 27B • Updated • 22.9k • 101
zai-org/GLM-OCR
Image-to-Text • Updated • 5.47M • • 1.59k
PaddlePaddle/PaddleOCR-VL-1.5
Image-Text-to-Text • 1.0B • Updated • 358k • 534
Jackrong/Qwopus3.5-9B-v3-GGUF
Image-Text-to-Text • 9B • Updated • 76.9k • 224
liked 2
models about 2 months ago
