reWordBench: Benchmarking and Improving the Robustness of Reward Models with Transformed Inputs

• URL: http://arxiv.org/abs/2503.11751v1
• Date: Fri, 14 Mar 2025 17:59:41 GMT
• Title: reWordBench: Benchmarking and Improving the Robustness of Reward Models with Transformed Inputs
• Authors: Zhaofeng Wu, Michihiro Yasunaga, Andrew Cohen, Yoon Kim, Asli Celikyilmaz, Marjan Ghazvininejad,
• Abstract summary: We show that state-of-the-art reward models suffer from substantial performance degradation even with minor input transformations.<n>We propose to explicitly train them to assign similar scores to paraphrases, and find that this approach also improves robustness to other distinct kinds of transformations.
• Score: 64.29893431743608
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Reward models have become a staple in modern NLP, serving as not only a scalable text evaluator, but also an indispensable component in many alignment recipes and inference-time algorithms. However, while recent reward models increase performance on standard benchmarks, this may partly be due to overfitting effects, which would confound an understanding of their true capability. In this work, we scrutinize the robustness of reward models and the extent of such overfitting. We build **reWordBench**, which systematically transforms reward model inputs in meaning- or ranking-preserving ways. We show that state-of-the-art reward models suffer from substantial performance degradation even with minor input transformations, sometimes dropping to significantly below-random accuracy, suggesting brittleness. To improve reward model robustness, we propose to explicitly train them to assign similar scores to paraphrases, and find that this approach also improves robustness to other distinct kinds of transformations. For example, our robust reward model reduces such degradation by roughly half for the Chat Hard subset in RewardBench. Furthermore, when used in alignment, our robust reward models demonstrate better utility and lead to higher-quality outputs, winning in up to 59% of instances against a standardly trained RM.

Related papers

• CARMO: Dynamic Criteria Generation for Context-Aware Reward Modelling [27.86204841898399]
  Reward modeling in large language models is susceptible to reward hacking.<n>We propose Context-Aware Reward Modeling (CARMO) to mitigate this problem.<n>We establish a new state-of-the-art performance in zero-shot settings for generative models, achieving a 2.1% improvement on Reward Bench.
  arXiv  Detail & Related papers  (2024-10-28T21:18:49Z)
• RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style [37.97757796124621]
  RM-Bench is a novel benchmark designed to evaluate reward models based on their sensitivity to subtle content differences and resistance to style biases. We evaluate nearly 40 reward models on RM-Bench and find that even state-of-the-art models achieve an average performance of only 46.6%.
  arXiv  Detail & Related papers  (2024-10-21T16:48:26Z)
• Evaluating Robustness of Reward Models for Mathematical Reasoning [14.97819343313859]
  We introduce a new design for reliable evaluation of reward models, and to validate this, we construct RewardMATH. We demonstrate that the scores on RewardMATH strongly correlate with the results of optimized policy and effectively estimate reward overoptimization.
  arXiv  Detail & Related papers  (2024-10-02T16:39:58Z)
• Scalable Ensembling For Mitigating Reward Overoptimisation [24.58937616758007]
  Reinforcement Learning from Human Feedback has enabled significant advancements within language modeling for powerful, instruction-following models. The alignment of these models remains a pressing challenge as the policy tends to overfit the learned proxy" reward model past an inflection point of utility.
  arXiv  Detail & Related papers  (2024-06-03T05:46:53Z)
• Confidence-aware Reward Optimization for Fine-tuning Text-to-Image Models [85.96013373385057]
  Fine-tuning text-to-image models with reward functions trained on human feedback data has proven effective for aligning model behavior with human intent. However, excessive optimization with such reward models, which serve as mere proxy objectives, can compromise the performance of fine-tuned models. We propose TextNorm, a method that enhances alignment based on a measure of reward model confidence estimated across a set of semantically contrastive text prompts.
  arXiv  Detail & Related papers  (2024-04-02T11:40:38Z)
• RewardBench: Evaluating Reward Models for Language Modeling [100.28366840977966]
  We present RewardBench, a benchmark dataset and code-base for evaluation of reward models. The dataset is a collection of prompt-chosen-rejected trios spanning chat, reasoning, and safety. On the RewardBench leaderboard, we evaluate reward models trained with a variety of methods.
  arXiv  Detail & Related papers  (2024-03-20T17:49:54Z)
• Model Reprogramming Outperforms Fine-tuning on Out-of-distribution Data in Text-Image Encoders [56.47577824219207]
  In this paper, we unveil the hidden costs associated with intrusive fine-tuning techniques. We introduce a new model reprogramming approach for fine-tuning, which we name Reprogrammer. Our empirical evidence reveals that Reprogrammer is less intrusive and yields superior downstream models.
  arXiv  Detail & Related papers  (2024-03-16T04:19:48Z)
• Helping or Herding? Reward Model Ensembles Mitigate but do not Eliminate Reward Hacking [62.146953368613815]
  Reward models play a key role in aligning language model applications towards human preferences. A natural mitigation is to train an ensemble of reward models, aggregating over model outputs to obtain a more robust reward estimate. We show that reward ensembles do not eliminate reward hacking because all reward models in the ensemble exhibit similar error patterns.
  arXiv  Detail & Related papers  (2023-12-14T18:59:04Z)
• QualEval: Qualitative Evaluation for Model Improvement [82.73561470966658]
  We propose QualEval, which augments quantitative scalar metrics with automated qualitative evaluation as a vehicle for model improvement. QualEval uses a powerful LLM reasoner and our novel flexible linear programming solver to generate human-readable insights. We demonstrate that leveraging its insights, for example, improves the absolute performance of the Llama 2 model by up to 15% points relative.
  arXiv  Detail & Related papers  (2023-11-06T00:21:44Z)
• Enhancing Certified Robustness via Smoothed Weighted Ensembling [7.217295098686032]
  We employ a Smoothed WEighted ENsembling scheme to improve the performance of randomized smoothed classifiers. We show the ensembling generality that SWEEN can help achieve optimal certified robustness. We also develop an adaptive prediction algorithm to reduce the prediction and certification cost of SWEEN models.
  arXiv  Detail & Related papers  (2020-05-19T11:13:43Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.