PulseAugur
实时 13:11:18
English(EN) Valid $\ne$ Necessary: Diagnosing Latent Inefficiency in Chain-of-Thought

新研究强调了大型语言模型和视觉语言模型中思维链的低效和过度自信

研究人员发现,大型语言模型(LLMs)的思维链(CoT)提示存在低效问题,其中有效但冗余的推理步骤会增加计算成本,但不会提高准确性。新开发的诊断基准 RIV-GSM8K 和一个名为 CAID 的指标,用于识别和惩罚这些“信息泡沫”步骤。一种事后压缩策略 PACE,利用 CAID,在保持准确性的同时,在各种基准测试中显著减少了 token(31-53%)。另外,值得注意的是,视觉语言模型(VLMs)中的 CoT 提示可能导致过度自信,这是因为不确定性估计依赖于模型自身的推理过程,而不是真实的不确定性。 AI

影响 识别出降低 LLM 和 VLM 推理计算成本和提高可靠性的方法,有望带来更高效、更值得信赖的 AI 系统。

排序理由 该集群包含两篇在 arXiv 上发表的学术论文,详细介绍了对 LLM 和 VLM 中思维链提示的低效和副作用的研究。

在 arXiv cs.AI 阅读 →

AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →

新研究强调了大型语言模型和视觉语言模型中思维链的低效和过度自信

报道来源 [3]

  1. arXiv cs.AI TIER_1 English(EN) · Daeyeop Lee, Hwanjo Yu ·

    Valid $\ne$ Necessary: Diagnosing Latent Inefficiency in Chain-of-Thought

    arXiv:2607.11266v1 Announce Type: new Abstract: Chain-of-Thought (CoT) prompting has significantly advanced the reasoning capabilities of Large Language Models (LLMs), yet it often incurs substantial computational costs due to over-reasoning: the generation of redundant, verbose,…

  2. arXiv cs.LG TIER_1 English(EN) · Robert Welch, Emir Konuk, Kevin Smith ·

    The Cost of Reasoning: Chain-of-Thought Induces Overconfidence in Vision-Language Models

    arXiv:2603.16728v2 Announce Type: replace Abstract: Vision-language models (VLMs) are increasingly deployed in high-stakes settings where reliable uncertainty quantification (UQ) is as important as predictive accuracy. Extended reasoning via chain-of-thought (CoT) prompting or re…

  3. arXiv cs.AI TIER_1 English(EN) · Hwanjo Yu ·

    有效不等于必要:诊断思维链中的潜在低效

    Chain-of-Thought (CoT) prompting has significantly advanced the reasoning capabilities of Large Language Models (LLMs), yet it often incurs substantial computational costs due to over-reasoning: the generation of redundant, verbose, or irrelevant steps. While existing reasoning s…