English(EN) Sample More, Reflect Less: Self-Refine and Reflexion Lose to Repeated Sampling at Equal Token Cost, from 1.5B to 7B

研究发现：大语言模型中的自我反思方法不敌简单的重复采样

作者 PulseAugur 编辑部 · [1 个来源] · 2026-07-31 04:00

一篇新发布的 arXiv 研究论文对大型语言模型中复杂的自我反思和精炼方法的有效性提出了质疑。研究人员发现，在 token 成本相同的情况下，诸如重复采样答案并选择最常见答案等更简单的技术，其表现与更复杂的精炼方法相当甚至更好。这一发现适用于从 1.5B 到 7B 参数的模型，并在数学基准测试中得到验证，表明自我检查的额外复杂性并不能可靠地提高准确性。 AI

影响表明更简单、更有效的方法可能更适合提高大语言模型的性能，从而可能降低计算成本。

排序理由学术论文，展示大语言模型方法的创新研究成果。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CL 阅读 →

1.5B
7B

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: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 学术论文，展示大语言模型方法的创新研究成果。[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: 60 days old; Aged out of breaking-news scoring windows; ranking reflects the durable signal from the full source set.

完整方法见我们的编辑标准。

报道来源 [1]

arXiv cs.CL TIER_1 English(EN) · Iliya Mirzaei · 2026-07-31 04:00

多采样，少反思：在相同代币成本下，自精炼和反思法输给重复采样，从1.5B到7B

arXiv:2607.28576v1 Announce Type: new Abstract: Methods that make a language model plan, criticise and rewrite its own answer, reflect on mistakes, pick the best of several attempts, or debate with copies of itself nearly all make it generate far more text than a single chain of …

本文如何被排名

报道来源 [1]

多采样，少反思：在相同代币成本下，自精炼和反思法输给重复采样，从1.5B到7B

相关话题