🚀NEW LABGetting Started with Claude AgentsStart lab
Evaluation · Reinforcement Learning

RewardBench 2

First page
RewardBench 2
Paper summary

RewardBench 2 is a new multi-skill benchmark for evaluating reward models with more challenging human prompts and stronger correlation to downstream performance. It highlights gaps in the current reward model's effectiveness and aims to support more rigorous evaluation, showing existing models score ~20 points lower than their predecessor.

Ask this paper

Every Monday
Get next week’s papers.
Subscribe on Substack