Think-RM: Enabling Long-Horizon Reasoning in Generative Reward Models

• URL: http://arxiv.org/abs/2505.16265v1
• Date: Thu, 22 May 2025 05:56:11 GMT
• Title: Think-RM: Enabling Long-Horizon Reasoning in Generative Reward Models
• Authors: Ilgee Hong, Changlong Yu, Liang Qiu, Weixiang Yan, Zhenghao Xu, Haoming Jiang, Qingru Zhang, Qin Lu, Xin Liu, Chao Zhang, Tuo Zhao,
• Abstract summary: Think-RM generates flexible, self-guided reasoning traces that support advanced capabilities.<n>Think-RM achieves state-of-the-art results on RM-Bench, outperforming both BT RM and vertically scaled GenRM by 8%.
• Score: 50.4652276723694
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement learning from human feedback (RLHF) has become a powerful post-training paradigm for aligning large language models with human preferences. A core challenge in RLHF is constructing accurate reward signals, where the conventional Bradley-Terry reward models (BT RMs) often suffer from sensitivity to data size and coverage, as well as vulnerability to reward hacking. Generative reward models (GenRMs) offer a more robust alternative by generating chain-of-thought (CoT) rationales followed by a final reward. However, existing GenRMs rely on shallow, vertically scaled reasoning, limiting their capacity to handle nuanced or complex (e.g., reasoning-intensive) tasks. Moreover, their pairwise preference outputs are incompatible with standard RLHF algorithms that require pointwise reward signals. In this work, we introduce Think-RM, a training framework that enables long-horizon reasoning in GenRMs by modeling an internal thinking process. Rather than producing structured, externally provided rationales, Think-RM generates flexible, self-guided reasoning traces that support advanced capabilities such as self-reflection, hypothetical reasoning, and divergent reasoning. To elicit these reasoning abilities, we first warm-up the models by supervised fine-tuning (SFT) over long CoT data. We then further improve the model's long-horizon abilities by rule-based reinforcement learning (RL). In addition, we propose a novel pairwise RLHF pipeline that directly optimizes policies using pairwise preference rewards, eliminating the need for pointwise reward conversion and enabling more effective use of Think-RM outputs. Experiments show that Think-RM achieves state-of-the-art results on RM-Bench, outperforming both BT RM and vertically scaled GenRM by 8%. When combined with our pairwise RLHF pipeline, it demonstrates superior end-policy performance compared to traditional approaches.

Related papers

• Good Learners Think Their Thinking: Generative PRM Makes Large Reasoning Model More Efficient Math Learner [31.033131727230277]
  Large reasoning models (LRMs) have recently shown promise in solving complex math problems when optimized with Reinforcement Learning (RL)<n>We propose a novel intrinsic signal-driven generative process evaluation mechanism operating at the thought level to address major bottlenecks in RL-based training.<n>Experiments on 1.5B and 7B parameter LRMs demonstrate that our method achieves higher problem-solving accuracy with significantly fewer training samples than outcome-only reward baselines.
  arXiv  Detail & Related papers  (2025-07-31T07:54:58Z)
• ReasonFlux-PRM: Trajectory-Aware PRMs for Long Chain-of-Thought Reasoning in LLMs [56.32212611983997]
  We introduce ReasonFlux-PRM, a novel trajectory-aware PRM to evaluate trajectory-response type of reasoning traces.<n>ReasonFlux-PRM incorporates both step-level and trajectory-level supervision, enabling fine-grained reward assignment aligned with structured chain-of-thought data.<n>Our derived ReasonFlux-PRM-7B yields consistent performance improvements, achieving average gains of 12.1% in supervised fine-tuning, 4.5% in reinforcement learning, and 6.3% in test-time scaling.
  arXiv  Detail & Related papers  (2025-06-23T17:59:02Z)
• Generative RLHF-V: Learning Principles from Multi-modal Human Preference [15.068452240642884]
  We introduce Generative RLHF-V, a novel alignment framework that integrates GRMs with multi-modal RLHF.<n>We propose a two-stage pipeline: $textbfmulti-modal generative reward modeling from RL$, where RL guides GRMs to actively capture human intention, then predict the correct pair-wise scores.<n>Our framework improves 4 MLLMs' performance across 7 benchmarks by $18.1%$, while the baseline RLHF is only $5.3%$.
  arXiv  Detail & Related papers  (2025-05-24T05:50:07Z)
• Reward Reasoning Model [104.39256985858428]
  Reward Reasoning Models (RRMs) are designed to execute a deliberate reasoning process before generating final rewards.<n>We implement a reinforcement learning framework that fosters self-evolved reward reasoning capabilities.<n> Notably, RRMs can adaptively exploit test-time compute to further improve reward accuracy.
  arXiv  Detail & Related papers  (2025-05-20T17:58:03Z)
• RM-R1: Reward Modeling as Reasoning [81.50471199906738]
  Reasoning Reward Models (ReasRMs) formulate reward modeling as a reasoning task.<n>We propose a reasoning-oriented training pipeline and train a family of ReasRMs, RM-R1.<n>Our models achieve state-of-the-art performance across three reward model benchmarks on average.
  arXiv  Detail & Related papers  (2025-05-05T06:11:12Z)
• Inference-Time Scaling for Generalist Reward Modeling [25.62000059973935]
  Reinforcement learning (RL) has been widely adopted in post-training for large language models (LLMs) at scale.<n>Key challenge of RL is to obtain accurate reward signals for LLMs in various domains beyond verifiable questions or artificial rules.<n>In this work, we investigate how to improve reward modeling with more inference compute for general queries.
  arXiv  Detail & Related papers  (2025-04-03T11:19:49Z)
• Towards Hierarchical Multi-Step Reward Models for Enhanced Reasoning in Large Language Models [33.547353090281284]
  We propose a novel reward model approach called the Hierarchical Reward Model.<n>It evaluates both individual and consecutive reasoning steps at both fine-grained and coarse-grained levels.<n>It excels at assessing multi-step reasoning coherence, especially when flawed steps are later corrected through self-reflection.
  arXiv  Detail & Related papers  (2025-03-16T15:18:40Z)
• WARM: On the Benefits of Weight Averaged Reward Models [63.08179139233774]
  We propose Weight Averaged Reward Models (WARM) to mitigate reward hacking. Experiments on summarization tasks, using best-of-N and RL methods, shows that WARM improves the overall quality and alignment of LLM predictions.
  arXiv  Detail & Related papers  (2024-01-22T18:27:08Z)
• Direct Preference Optimization: Your Language Model is Secretly a Reward Model [119.65409513119963]
  We introduce a new parameterization of the reward model in RLHF that enables extraction of the corresponding optimal policy in closed form. The resulting algorithm, which we call Direct Preference Optimization (DPO), is stable, performant, and computationally lightweight. Our experiments show that DPO can fine-tune LMs to align with human preferences as well as or better than existing methods.
  arXiv  Detail & Related papers  (2023-05-29T17:57:46Z)

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.