arXiv:2509.05489cs.LG2025-09被引 8

用自对齐奖励提升大模型推理准确率与效率

Self-Aligned Reward: Towards Effective and Efficient Reasoners

  • 基于困惑度差值设计自对齐奖励,鼓励简洁且紧扣问题的回答
  • 在7个基准上提升准确率4%,推理成本降低30%
  • 适合追求高效精准推理的模型训练场景

基于可验证奖励的强化学习显著提升了大语言模型的推理能力,但此类信号仍较粗粒度,仅提供正确性二元反馈。这导致推理冗长、计算成本高,现有方案常以牺牲准确率为代价。为此,我们提出自对齐奖励(SAR),一种自引导信号,可补充可验证奖励,兼顾推理准确率与效率。SAR定义为在查询条件下的答案与独立答案之间的相对困惑度差异,从而偏好简洁且与查询相关的回应。定量分析显示,SAR能可靠区分答案质量:简洁正确的答案得分更高,部分正确高于完全错误。在4个模型、7个基准上的评估表明,将SAR与主流强化学习算法(如PPO和GRPO)结合,准确率提升4%,推理成本降低30%。进一步分析表明,相较于基于长度或自信心的奖励信号,SAR实现了正确性与效率的帕累托最优平衡。同时,SAR能在保留高级推理行为的前提下缩短响应长度,有效抑制无意义扩展。这些结果凸显了自对齐奖励作为细粒度补充信号的潜力,为更高效、有效的大模型训练开辟新路径。

原文摘要 · Abstract (English)

Reinforcement learning with verifiable rewards has significantly advanced reasoning in large language models (LLMs), but such signals remain coarse, offering only binary correctness feedback. This limitation often results in inefficiencies, including overly verbose reasoning and high computational cost, while existing solutions often compromise accuracy. To address this, we introduce self-aligned reward (SAR), a self-guided signal that complements verifiable rewards to encourage both reasoning accuracy and efficiency. SAR is defined as the relative perplexity difference between an answer conditioned on the query and the standalone answer, thereby favoring responses that are concise and query-specific. Quantitative analysis reveals that SAR reliably distinguishes answer quality: concise, correct answers score higher than redundant ones, and partially correct answers score higher than entirely incorrect ones. Evaluation on 4 models across 7 benchmarks shows that integrating SAR with prevalent RL algorithms like PPO and GRPO improves accuracy by 4%, while reducing inference cost by 30%. Further analysis demonstrates that SAR achieves a Pareto-optimal trade-off between correctness and efficiency compared to reward signals based on length or self-confidence. We also show that SAR shortens responses while preserving advanced reasoning behaviors, demonstrating its ability to suppress unnecessary elaboration without losing critical reasoning. These results highlight the promise of self-aligned reward as a fine-grained complement to verifiable rewards, paving the way for more efficient and effective LLM training.

强化学习推理优化大模型

Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。