Aug 2026· Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2· pp. 9730-9741· 0 citations· 20 references
Abstract
Reward Models (RMs) are widely used to guide the alignment of Large Language Models (LLMs) in Reinforcement Learning from Human Feedback (RLHF), and their quality has a substantial impact on the performance of trained policy models. In current practice, reward models are predominantly evaluated using local metrics such as pairwise accuracy on validation sets. However, recent studies have shown that the accuracy metric exhibits only weak correlation with downstream policy performance. To bridge this gap, we propose RMEval, a benchmark that revisits reward model evaluation through the lens of global ranking consistency. Specifically, we conduct the standard RLHF training to generate a sequence of policy checkpoints with progressively improving performance. We then construct our evaluation sets by sampling responses from these checkpoints, establishing a ground-truth ranking based on downstream policy evaluation. By scoring these responses with a candidate reward model, we measure the consistency between the ranking induced by the reward model and the ground-truth ranking defined by policy performance, using the consistency as our evaluative metric. Through extensive empirical analysis, we evaluate existing reward models on RMEval and demonstrate that our metric exhibits a strong correlation with downstream policy performance, with Pearson coefficients exceeding 0.6 across multiple domains.
This work proposes LaRes, a novel hybrid framework that achieves efficient policy learning through reward function search by leveraging large language models to generate the reward function population, guiding RL in policy learning.
Pengyi Li, Hongyao Tang, Jinbin Qiao et al.· Neural Information Processin...· 4 citations
Ranking is a fundamental component of modern information access systems. Reinforcement learning (RL) provides a flexible framework for directly optimizing coarse-grained feedback and system-level objectives defined over the complete ranking list. However, existing RL-based ranking methods typically treat each sampled permutation as an atomic output and evaluate it primarily through a scalar reward, overlooking the structural relationships among different ranking lists. Consequently, permutations with similar rewards but substantially different permutation patterns may receive comparable optimization signals, potentially leading to inaccurate credit assignment and overly aggressive policy updates. To address this limitation, we propose SRPO, a \textbf{S}tructure-aware \textbf{R}elative \textbf{P}olicy \textbf{O}ptimization framework for listwise ranking. SRPO measures the discrepancy between sampled permutations using a top-weighted Kendall-tau distance and normalizes their pairwise reward differences by the corresponding distances. It quantifies the reward improvement per unit of ranking change, thereby emphasizing efficient local refinements, particularly those involving top-ranked positions. Experimental results across two ranking scenarios demonstrate that explicitly modeling permutation-level differences improves the effectiveness and stability of listwise ranking, with particularly favorable performance in limited-feedback and complex list-level optimization settings.
Yiteng Tu, Weihang Su, Zitao Su et al.· arXiv.org· 0 citations
Recent advances in reward modeling show a paradigm shift from discriminative reward models to generative reward models. However, despite their strong capabilities in response ranking, generative reward models have not realized their potential in reinforcement learning (RL). Our analysis reveals that this limitation arises from a mismatch between the comparative nature of generative reward modeling and the scalar scoring paradigm adopted by existing RL algorithms. To bridge this gap, we propose a Ranking-based Reward Construction (RRC) approach, which enables generative reward models to provide more effective RL learning signals by deriving rewards from relative preference rankings. RRC introduces two complementary strategies: self-competitive ranking, which exploits comparisons among sampled responses, and anchor-guided ranking, which enables scalable ranking-based reward construction with a small set of reference responses. Experiments across open-ended chat and reasoning benchmarks demonstrate that RRC substantially improves RL training with generative reward models, achieving consistent gains over existing reward construction approaches. Our code can be found at https://github.com/wangclnlp/RRC.
Chenglong Wang, Ziming Zhu, Yifu Huo et al.· 0 citations
Reinforcement learning (RL) has become one of the primary paradigms for reasoning enhancement of large language models (LLMs). In particular, Group Relative Policy Optimization (GRPO) and related algorithms have demonstrated strong performance with outcome-level rewards. However, these methods depend solely on the final answer, without feedback regarding which intermediate steps contribute to success or failure. As task complexity and reasoning trajectory length increase, such sparse final-answer rewards become increasingly insufficient. To address this limitation, we introduce ConsensusBench, a novel dataset designed to provide rule-based process-level signals. We posit that a correct final answer relies on a small set of intermediate conclusions throughout the reasoning process, which can be seen as a verifiable sub-outcome. We identify these sub-outcomes by filtering correct trajectories from N rollouts and clustering semantically equivalent intermediate statements. We call these clustered statements as Consensus Nodes. By integrating a rule-based process reward derived from these nodes into GRPO-style algorithms, we develop a new reinforcement learning signal named ConsensusPR. It directly reduces the reward sparsity of outcome reward across long reasoning trajectories. To facilitate systematic process-level evaluation, we introduce three metrics to our benchmark: Final Answer Accuracy (Acc), Node Coverage Rate (NCR), and Tokens per Node (TPN). Experiments across AIME 2024, AIME 2025, GSM8K, MATH-500, and our ConsensusBench demonstrate that the proposed method consistently surpasses GRPO-style approaches, highlighting the practical value of consensus nodes in guiding reasoning.
Shi-Qi Yan, Chao-Hong Tan, Qian Chen et al.· 0 citations
Data policies for reinforcement learning with verifiable rewards (RLVR) determine which rollouts are used, how strongly they are weighted, and which domains contribute to subsequent training batches. We introduce DataFlex-RL, an evaluation platform for comparing these choices under a common GRPO recipe. Our primary experiment evaluates 13 configurations across 12 matched seeds using Qwen2.5-7B-Base and 12 mathematics, logic, and science benchmarks. Uniform GRPO improves the domain-balanced average accuracy by 7.76 percentage points over the untrained checkpoint. None of the eight rollout-selection or reweighting methods achieves a paired 95% confidence interval that excludes zero relative to uniform sampling, and none of the three adaptive mixtures outperforms a fixed equal mixture at the same level of precision. A corrected 12-seed extension on Llama-3.1-8B-Base places the additional methods on the same score scale as the original controls, but does not reveal a consistent winner in terms of observed mean performance. We also quantify evaluation sensitivity by rescoring nine Qwen2.5-7B-Instruct runs using a math-heavy six-benchmark summary, consisting of five mathematics benchmarks and GPQA-Diamond but no logic benchmark, and comparing it with the domain-balanced 12-benchmark summary. The resulting rankings are negatively correlated, with a correlation coefficient of -0.33, whereas summaries that retain all 12 benchmarks largely agree. Across the controlled settings studied here, changing the data policy measurably changes the training process but does not produce a reproducible improvement over uniform training.
Hao Liang, Mingrui Chen, Hengyi Feng et al.· 0 citations
Reinforcement learning from verifiable rewards (RLVR) drives chain-of-thought reasoning in large language models, yet its binary outcome reward cannot distinguish among correct trajectories. Existing dense reward alternatives, from surface heuristics to process reward models, either ignore the expert solutions already present in training corpora or require expensive offline annotation. We propose Gradient-Aligned Reward (GAR), which operates in the policy's own gradient space: truncated backpropagation through the output projection layer extracts a compact gradient vector for each rollout, and cosine similarity with an expert-anchor gradient yields a dense, reasoning-aware reward with less than 9% wall-clock overhead. We prove that this cosine admits a multiplicative decomposition into prediction-error and activation-pattern factors, providing a concrete characterization of what the alignment signal measures. On Qwen3-4B and Qwen3-8B, GAR consistently improves over GRPO and other baselines on competition-level math benchmarks and transfers to GPQA Diamond and MMLU-Pro without domain-specific data. Code and data are available at https://github.com/LQgdwind/GAR.
Le-Qi Zheng, Jin-Bo Su, Fang Niu et al.· 2 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
