Process Reward Models: The Navigator for LLM Reasoning
More inference compute is not enough on its own. This article explains how PRMs score reasoning steps, rerank Best-of-N candidates, and guide search.
Latest notes
Recent posts
Building an AI Stock Agent That Reads My Portfolio and Debates the Next Market Session
When you own stocks, your thoughts do not stop just because the market has closed.
Why Natural-Language Generation Prompts Are So Hard to Improve
There are many ways to use LLMs. This article focuses on tasks that generate natural language shown directly to users.
