Rectify Evaluation Preference: Improving LLMs' Critique on Math Reasoning via Perplexity-aware Reinforcement Learning

• URL: http://arxiv.org/abs/2511.10303v1
• Date: Fri, 14 Nov 2025 01:44:32 GMT
• Title: Rectify Evaluation Preference: Improving LLMs' Critique on Math Reasoning via Perplexity-aware Reinforcement Learning
• Authors: Changyuan Tian, Zhicong Lu, Shuang Qian, Nayu Liu, Peiguang Li, Li Jin, Leiyi Hu, Zhizhao Zeng, Sirui Wang, Ke Zeng, Zhi Guo,
• Abstract summary: In this paper, we quantify and investigate the potential reason -- imbalanced evaluation preference.<n>Motivated by the analysis of the reason, a novel perplexity-aware reinforcement learning algorithm is proposed to rectify the evaluation preference.
• Score: 34.43632129774481
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: To improve Multi-step Mathematical Reasoning (MsMR) of Large Language Models (LLMs), it is crucial to obtain scalable supervision from the corpus by automatically critiquing mistakes in the reasoning process of MsMR and rendering a final verdict of the problem-solution. Most existing methods rely on crafting high-quality supervised fine-tuning demonstrations for critiquing capability enhancement and pay little attention to delving into the underlying reason for the poor critiquing performance of LLMs. In this paper, we orthogonally quantify and investigate the potential reason -- imbalanced evaluation preference, and conduct a statistical preference analysis. Motivated by the analysis of the reason, a novel perplexity-aware reinforcement learning algorithm is proposed to rectify the evaluation preference, elevating the critiquing capability. Specifically, to probe into LLMs' critiquing characteristics, a One-to-many Problem-Solution (OPS) benchmark is meticulously constructed to quantify the behavior difference of LLMs when evaluating the problem solutions generated by itself and others. Then, to investigate the behavior difference in depth, we conduct a statistical preference analysis oriented on perplexity and find an intriguing phenomenon -- ``LLMs incline to judge solutions with lower perplexity as correct'', which is dubbed as \textit{imbalanced evaluation preference}. To rectify this preference, we regard perplexity as the baton in the algorithm of Group Relative Policy Optimization, supporting the LLMs to explore trajectories that judge lower perplexity as wrong and higher perplexity as correct. Extensive experimental results on our built OPS and existing available critic benchmarks demonstrate the validity of our method.

Related papers

• Rectifying LLM Thought from Lens of Optimization [48.98086817378953]
  Long chain-of-thought (CoT) prompting enables thorough exploration and deliberation.<n>Despite advances, long-CoT LLMs often exhibit suboptimal reasoning behaviors.<n>We introduce RePro, a novel approach to refine LLM reasoning during post-training.
  arXiv  Detail & Related papers  (2025-12-01T17:41:08Z)
• IF-CRITIC: Towards a Fine-Grained LLM Critic for Instruction-Following Evaluation [87.38454788767545]
  We propose IF-CRITIC, an evaluation model for instruction following in Large Language Models.<n>With the scalable reward signals provided by IF-CRITIC, LLMs can achieve substantial performance gains in instruction-following optimization.
  arXiv  Detail & Related papers  (2025-11-02T17:06:49Z)
• What Defines Good Reasoning in LLMs? Dissecting Reasoning Steps with Multi-Aspect Evaluation [67.47463575774388]
  We decompose reasoning quality into two dimensions: relevance and coherence.<n>To measure these aspects reliably, we introduce causal stepwise evaluation (CaSE)<n>We show that curating training data with CaSE-evaluated relevance and coherence directly improves final task performance.
  arXiv  Detail & Related papers  (2025-10-23T14:30:37Z)
• Reasoning with Preference Constraints: A Benchmark for Language Models in Many-to-One Matching Markets [13.111181135818184]
  Large language models (LLMs) have shown strong performance on complex mathematical tasks, including optimization.<n>Applying LLMs to matching problems, which require reasoning under preferential and structural constraints, remains underexplored.<n>We employ a novel benchmark of 369 instances of the College Admission Problem to evaluate LLMs across key dimensions: feasibility, stability, and optimality.
  arXiv  Detail & Related papers  (2025-09-16T14:48:46Z)
• Supervised Optimism Correction: Be Confident When LLMs Are Sure [91.7459076316849]
  We establish a novel theoretical connection between supervised fine-tuning and offline reinforcement learning.<n>We show that the widely used beam search method suffers from unacceptable over-optimism.<n>We propose Supervised Optimism Correction, which introduces a simple yet effective auxiliary loss for token-level $Q$-value estimations.
  arXiv  Detail & Related papers  (2025-04-10T07:50:03Z)
• Self-Evolving Critique Abilities in Large Language Models [59.861013614500024]
  This paper explores enhancing critique abilities of Large Language Models (LLMs)<n>We introduce SCRIT, a framework that trains LLMs with self-generated data to evolve their critique abilities.<n>Our analysis reveals that SCRIT's performance scales positively with data and model size.
  arXiv  Detail & Related papers  (2025-01-10T05:51:52Z)
• Critic-CoT: Boosting the reasoning abilities of large language model via Chain-of-thoughts Critic [48.94340387130627]
  Critic-CoT is a framework that pushes LLMs toward System-2-like critic capability.<n>CoT reasoning paradigm and the automatic construction of distant-supervision data without human annotation.<n>Experiments on GSM8K and MATH demonstrate that our enhanced model significantly boosts task-solving performance.
  arXiv  Detail & Related papers  (2024-08-29T08:02:09Z)
• MR-Ben: A Meta-Reasoning Benchmark for Evaluating System-2 Thinking in LLMs [55.20845457594977]
  Large language models (LLMs) have shown increasing capability in problem-solving and decision-making.<n>We present a process-based benchmark MR-Ben that demands a meta-reasoning skill.<n>Our meta-reasoning paradigm is especially suited for system-2 slow thinking.
  arXiv  Detail & Related papers  (2024-06-20T03:50:23Z)

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.