English(EN) Taking the Longest Branch Scores 76.2% Against CoT-Decoding's 75.9%, So the Logprobs Bought Nothing

最长分支方法在LLM响应选择中优于CoT-Decoding

作者 PulseAugur 编辑部 · [1 个来源] · 2026-08-21 15:38

最近的一项实验探索了从大型语言模型中选择最佳响应的不同方法，将简单的“最长分支”方法与更复杂的CoT-Decoding技术进行了比较。实验发现，“最长分支”方法（在不使用对数概率的情况下选择具有最多token的分支）的准确率达到了76.2%，高于CoT-Decoding的75.9%。这表明涉及对数概率访问和答案跨度定位的复杂方法，与更简单的自由控制方法相比，并未显著提高性能。 AI

影响表明对于某些LLM任务，更简单的提示策略可能与复杂策略一样有效。

排序理由该条目描述了一项比较不同LLM提示技术的新实验发现。[lever_c_demoted from research: ic=1 ai=1.0]

在 dev.to — LLM tag 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: 0 / 100; Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket: Tool; 该条目描述了一项比较不同LLM提示技术的新实验发现。[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, other; 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: 48 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]

dev.to — LLM tag TIER_1 English(EN) · Devanshu Biswas · 2026-08-21 15:38

Longest Branch 准确率达 76.2%，CoT-Decoding 为 75.9%，Logprobs 毫无用处

Sixty-six entries in this series changed the prompt. CoT-decoding (Wang & Zhou, 2024) changes nothing about it. Branch the top-k first tokens instead of taking the argmax one, greedy-decode each, and several of those branches turn out to contain a…

本文如何被排名

报道来源 [1]

Longest Branch 准确率达 76.2%，CoT-Decoding 为 75.9%，Logprobs 毫无用处

相关实体

相关话题