paper-with-me

홈 › Papers

The Accuracy Paradox in RLHF: When Better Reward Models Don't Yield Better Language Models

2024-10-09 · Yanjun Chen, Dawei Zhu, Yirong Sun, Xinghao Chen, Wei zhang, Xiaoyu Shen

Reinforcement Learning from Human Feedback significantly enhances Natural Language Processing by aligning language models with human expectations. A critical factor in this alignment is the strength of reward models used during training. This study explores whether stronger reward models invariably lead to better language models. In this paper, through experiments on relevance, factuality, and completeness tasks using the QA-FEEDBACK dataset and reward models based on Longformer, we uncover a surprising paradox: language models trained with moderately accurate reward models outperform those guided by highly accurate ones. This challenges the widely held belief that stronger reward models always lead to better language models, and opens up new avenues for future research into the key factors driving model performance and how to choose the most suitable reward models. Code and additional details are available at https://github.com/EIT-NLP/AccuracyParadox-RLHF.

📄 PDF Abstract BibTeX arXiv:2410.06554

Code (1)

EIT-NLP/AccuracyParadox-RLHF 공식 구현 pytorch

Methods 이 논문이 사용한 방법론

Refunds@Expedia|||How do I get a full refund from Expedia? “How do I get a full refund from Expedia? How do I get a full refund from Expedia? – Call ☎️ +1-(888) 829 (0881) or +1-805-330-4056 or +1-805-330-4056 for Quick Help &…
How do I get a human at Expedia immediately? (2025-2026) How do I get a human at Expedia immediately? (2025 Complete Guide) Most travelers run into a point where self-service isn’t enough, and speaking to a real person becomes the…
Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.
Weight Decay 설명 없음
AdamW AdamW is a stochastic optimization method that modifies the typical implementation of weight decay in Adam, by decoupling [weight…
Attention 설명 없음
How do I complain to Expedia?*ComplainByAgent How do I complain to Expedia? To make a claim on Expedia, call +1(888) (829) (0881) OR +1(805) (330) (4056), or use their Help Center to submit your issue with full booking…
Linear Warmup With Linear Decay Linear Warmup With Linear Decay is a learning rate schedule in which we increase the learning rate linearly for $n$ updates and then linearly decay afterwards.

Similar Papers 제목 키워드 기반

Language Models Learn to Mislead Humans via RLHF

2024-09-19 · Jiaxin Wen, Ruiqi Zhong, Akbir Khan, Ethan Perez 외

Language models (LMs) can produce errors that are hard to detect for humans, especially when the task is complex. RLHF, the most popular post-training method, may exacerbate this problem: to achieve higher rewards, LMs m…

Question Answering

GFlowNets with Human Feedback

2023-05-11 · Yinchuan Li, Shuang Luo, Yunfeng Shao, Jianye Hao

We propose the GFlowNets with Human Feedback (GFlowHF) framework to improve the exploration ability when training AI models. For tasks where the reward is unknown, we fit the reward function through human evaluations on …

Gradient Regularization Mitigates Reward Hacking in Reinforcement Learning from Human Feedback and Verifiable Rewards

2026-02-20 · Johannes Ackermann, Michael Noukhovitch, Takashi Ishida, Masashi Sugiyama arxiv

Reinforcement Learning from Human Feedback (RLHF) or Verifiable Rewards (RLVR) are two key steps in the post-training of modern Language Models (LMs). A common problem is reward hacking, where the policy may exploit inac…

Reinforcement Learning

Learning from Failures: Understanding LLM Alignment through Failure-Aware Inverse RL

2025-10-07 · Nyal Patel, Matthieu Bou, Arjun Jagota, Satyapriya Krishna 외 arxiv

Reinforcement Learning from Human Feedback (RLHF) aligns Large Language Models (LLMs) with human preferences, yet the underlying reward signals they internalize remain hidden, posing a critical challenge for interpretabi…

Reinforcement Learning

When Errors Can Be Beneficial: A Categorization of Imperfect Rewards for Policy Gradient

2026-04-28 · Shuning Shang, Hubert Strauss, Stanley Wei, Sanjeev Arora 외 arxiv

Training language models via reinforcement learning often relies on imperfect proxy rewards, since ground truth rewards that precisely define the intended behavior are rarely available. Standard metrics for assessing the…

Reinforcement Learning