新的ReAPO方法通过感知重试的信用分配来提升LLM推理能力

作者 PulseAugur 编辑部 · [1 个来源] · 2026-10-22 04:00

研究人员开发了一种名为重试感知策略优化（ReAPO）的新方法，以增强大型语言模型（LLM）的推理能力。该方法通过引入基于随机最短路径的新型信用分配框架，解决了LLM推理中探索效率低下的问题。ReAPO修改了现有的GRPO算法，使其能够更有效地从失败的尝试中学习，在不无意中奖励先前失败的情况下强化成功的重试。在Qwen3和OLMo-2模型上的各种基准测试实验表明，与其它方法相比，ReAPO提高了首次尝试的性能，并以显著更少的生成token达到了峰值准确率。 AI

影响通过优化训练过程中的探索和信用分配，提高了LLM的推理效率和性能。

排序理由该集群包含一篇详细介绍改进LLM推理新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.LG 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

本文如何被排名

Signal score: 1 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该集群包含一篇详细介绍改进LLM推理新方法的学术论文。[lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration: Single-source cluster; Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics: paper, model release; Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance: High; Clearly on-topic for AI-industry coverage.
Story freshness: 1 days old; Coverage has settled into its steady-state source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.LG TIER_1 English(EN) · Nghi Dao, Calder Morton-Ferguson, Noah Golowich, Adam Block · 2026-10-22 04:00

ReAPO: Retry-Aware Policy Optimization for Efficient Exploration in LLM Reasoning

arXiv:2610.28770v1 Announce Type: new Abstract: RLVR improves language-model reasoning, but commonly relies on independent stochastic rollouts for exploration, which can repeatedly sample similar failed strategies. Prior work has seen success with approaches that guide exploratio…

本文如何被排名

报道来源 [1]

ReAPO: Retry-Aware Policy Optimization for Efficient Exploration in LLM Reasoning

相关实体

相关话题