Get the latest tech news

Batched reward model inference and Best-of-N sampling


test

Reward models have been a key part of reinforcement learning on top of LLMs, used broadly in techniques like RLHF and as LLM-as-a-judge critics in evals. Since my cloud GPU bill this month is kind of ridiculous, I decided to evaluate using a small random 100 question subset of TruthfulQA, a multiple choice benchmark for LLMs. Best-of-N-Sampling doesn't add much latency (we can sample and score hundreds of completions in parallel with batching) and is super easy to implement 7.

Get the Android app

Or read this on Hacker News

Read more on:

Photo of sampling

sampling