Stories by m-a-r-c-e-l

AI Philosophy Competition

Diffusion as a Training Curriculum for Timestep-Free Iterative Reasoning

Unlocking Lossless Speedups in LLMs via Discrete Diffusion (5000 Tk/S)

Show HN: a Rust compiler translating Joern CPGs to Nuclei templates

Matching Features, Not Tokens: Energy-Based Fine-Tuning of Language Models

The Principle Benchmark

Language Models Can Control Their Own Attention

Inverse Rendering for Modeling with Line Primitives (Good for Fur)

Reverse Polish Notation Calculators You Can Still Buy

A Proof of Conway's Refinement Conjecture

Show HN: Rust compiler converting CPGs to Nuclei YAML

Built a GitHub Runners Benchmarking Website

No-Code Self-Documenting Makefiles

Bootstrapping enables human-level concept learning in self-supervised models

Reward Guided Speculative Decoding

Renderformer V2 (Full neural rendering)

DIY Gene Gun [video]

Beyond Booleans

Himalayan Monal

Why not to use the Gaussian kernel

1 2 3 4 5 6 7 8 9 10