How this site works: the work explorer and AI assistant
A practical guide to finding relevant work, understanding the connections on the map, and checking the sources behind the site's AI answers.
Read articleWriting
Clear explanations of LLM research, retrieval systems, and the decisions behind them.
Open-weight refusal is one line away from off. Here is how we made it cost more.
Refusal in instruction-tuned LLMs lives in a single direction you can delete with a rank-1 edit. We post-trained Llama-3.2-1B-Instruct to spread that signal across many directions, raising the attack rank from K=1 to K≥16 with no measurable utility hit.
Read articleNeurIPS 2025 Best Paper: Scaling Depth in Self-Supervised Reinforcement Learning
A structured academic recap of a NeurIPS 2025 Best Paper Award–winning work that studies how reinforcement learning can scale by adopting self-supervised objectives, supplemented with insights from the authors’ interview.
Read article