10 articles

📰 Dev.to · felipe muniz

Articles from Dev.to · felipe muniz · 10 articles · Updated every 3 hours · View all reads

All ⚡ AI Lessons (9465) ArXiv cs.AIDev.to · FORUM WEBForbes InnovationDev.to AIOpenAI NewsHugging Face Blog
DRM-Transformer — Intrinsic Geometry for Structural Alignment
Dev.to · felipe muniz 2w ago
DRM-Transformer — Intrinsic Geometry for Structural Alignment
Why don't current LLMs geometrically distinguish between saving and destroying...
Read article → + Playlist
We Found Toroidal Topology Emerging in a Neural Epistemic Manifold
Dev.to · felipe muniz 3w ago
We Found Toroidal Topology Emerging in a Neural Epistemic Manifold
Preliminary evidence that the 5D epistemic space of AletheionV2 converges toward toroidal topology as...
Read article → + Playlist
Encoding Human Values as Geometry: The Gravitational Objective
Dev.to · felipe muniz 3w ago
Encoding Human Values as Geometry: The Gravitational Objective
Part 2 of the AletheionLLM-v2 geometry series. Part 1: How to measure whether your model's...
Read article → + Playlist
How to Measure Whether Your Model's Uncertainty Space Is Flat or Curved
Dev.to · felipe muniz 3w ago
How to Measure Whether Your Model's Uncertainty Space Is Flat or Curved
A practical guide to Riemannian epistemic geometry in language models, with code. Most...
Read article → + Playlist
The Three Doors Problem: Why RLHF Systems Slide Toward Autonomy
Dev.to · felipe muniz 4w ago
The Three Doors Problem: Why RLHF Systems Slide Toward Autonomy
What happens when an AI detects it's lying to please you? Every AI trained with RLHF lives a silent...
Read article → + Playlist
The Manifold Game
Dev.to · felipe muniz 1mo ago
The Manifold Game
What if the AI you use knew — geometrically — what it doesn't know? Today I published "The Manifold...
Read article → + Playlist
I trained a 354M LLM alone and it outperforms GPT-2 Medium in epistemic calibration
Dev.to · felipe muniz 1mo ago
I trained a 354M LLM alone and it outperforms GPT-2 Medium in epistemic calibration
No team. No institutional funding. No university affiliation. Just me, a RunPod account with 5x...
Read article → + Playlist
The Semantic Gatekeeping Problem in AI Research
Dev.to · felipe muniz 1mo ago
The Semantic Gatekeeping Problem in AI Research
In 1847, Ignaz Semmelweis discovered that doctors washing their hands could reduce maternal...
Read article → + Playlist
ATIC Doesn't Train. It Thinks. — How a Brazilian Developer Hit #1 on LiveBench Without Touching a Single Weight
Dev.to · felipe muniz 1mo ago
ATIC Doesn't Train. It Thinks. — How a Brazilian Developer Hit #1 on LiveBench Without Touching a Single Weight
"More human than human." That was the motto of the Tyrell Corporation in Blade Runner....
Read article → + Playlist
How I Built a System to Detect When LLMs Don't Know Something
Dev.to · felipe muniz 1mo ago
How I Built a System to Detect When LLMs Don't Know Something
How I Built a System to Detect When LLMs Don't Know Something The Problem I Was...
Read article → + Playlist