Tagged with

3 articles found

The 1B-Parameter Insurrection: Maincoder-1B’s 76% HumanEval Score Challenges AI’s Bigger-Is-Better Orthodoxy

A tiny Australian startup just proved that scaling laws aren’t destiny. Their 1B-parameter coding model achieves best-in-class performance on HumanEval, forcing a reckoning with our obsession over parameter counts.

#benchmarks#code-generation#language-models...

ai-transparency

Olmo 3 Just Exposed How ‘Open’ Your Favorite AI Model Really Was

AI2’s latest release isn’t just another open-weight model, it’s a fully transparent AI system that challenges the industry’s definition of ‘open’ and reshapes the US-China AI race.

#ai-transparency#language-models#open-source-ai

bert

Diffusion Language Models Are Breaking the Autoregressive Dungeon

New framework turns BERT into a diffusion-based chatbot using discrete diffusion, and it’s rewriting how we think about language generation

#bert#diffusion#language-models...