REVIEW 6 cited by
RMB: Comprehensively Benchmarking Reward Models in LLM Alignment
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Reward models (RMs) guide the alignment of large language models (LLMs), steering them toward behaviors preferred by humans. Evaluating RMs is the key to better aligning LLMs. However, the current evaluation of RMs may not directly correspond to their alignment performance due to the limited distribution of evaluation data and evaluation methods that are not closely related to alignment objectives. To address these limitations, we propose RMB, a comprehensive RM benchmark that covers over 49 real-world scenarios and includes both pairwise and Best-of-N (BoN) evaluations to better reflect the effectiveness of RMs in guiding alignment optimization. We demonstrate a positive correlation between our benchmark and the downstream alignment task performance. Based on our benchmark, we conduct extensive analysis on the state-of-the-art RMs, revealing their generalization defects that were not discovered by previous benchmarks, and highlighting the potential of generative RMs. Furthermore, we delve into open questions in reward models, specifically examining the effectiveness of majority voting for the evaluation of reward models and analyzing the impact factors of generative RMs, including the influence of evaluation criteria and instructing methods. Our evaluation code and datasets are available at https://github.com/Zhou-Zoey/RMB-Reward-Model-Benchmark.
Forward citations
Cited by 6 Pith papers
-
Personalized RewardBench: Evaluating Reward Models with Human Aligned Personalization
Personalized RewardBench reveals that state-of-the-art reward models reach only 75.94% accuracy on personalized preferences and shows stronger correlation with downstream BoN and PPO performance than prior benchmarks.
-
AI Can Learn Scientific Taste
Reinforcement learning on citation-preference pairs teaches a model to predict which papers will be cited more and to propose ideas that LLM judges rate as likely to be cited more—but "taste" here means citation impact.
-
URPO: A Unified Reward & Policy Optimization Framework for Large Language Models
A seven-billion-parameter model trained with one unified GRPO loop improves instruction following, reasoning, and reward modeling at the same time.
-
CompassJudger-2: Towards Generalist Judge Model via Verifiable Rewards
A 7B judge model trained with verifiable reward signals and a margin contrastive loss matches the judgment accuracy of models tens of times larger, and a new benchmark JudgerBenchV2 standardizes judge evaluation.
-
Agent-RewardBench: Towards a Unified Benchmark for Reward Modeling across Perception, Planning, and Safety in Real-World Multimodal Agents
A new step-level reward-modeling benchmark for multimodal agents shows current MLLMs reach at most 61.6 percent accuracy, and benchmark score correlates strongly (r=0.981 across five models) with downstream A* search ...
-
Reward Modeling for Reinforcement Learning-Based LLM Reasoning: Design, Challenges, and Evaluation
A taxonomy-driven survey arguing that reward design is the central mechanism shaping reliable LLM reasoning, with maps of reward paradigms, reward-hacking failure modes, and benchmark pitfalls.
Discussion (0). Continue with ORCID to comment.