一项新的研究论文调查了小型语言模型（SLMs）的自我纠正能力，发现即使在提供了正确答案和提示的情况下，它们在改进推理方面仍然存在困难。该研究开发了一个三步流程来测试 SLMs 在算术和逻辑推理方面的能力，结果显示，在纠正性反馈下准确率仅提高了 4.4%。有趣的是，研究还表明，更长的提示有时会阻碍性能，这表明对于 SLMs 来说，增加的审议并不总是能带来更好的结果。 AI
影响 SLMs 表现出显著的自我纠正差距，这表明当前的架构可能需要根本性的改变才能实现强大的推理能力。
排序理由 该集群包含一篇详细介绍人工智能模型能力实验结果的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.08471v1 Announce Type: cross Abstract: Recently, language models have made rapid progress across various domains and applications. However, their capability for self-improvement, i.e., whether they are adept at recognising and correcting flaws in their own reasoning, r…
Recently, language models have made rapid progress across various domains and applications. However, their capability for self-improvement, i.e., whether they are adept at recognising and correcting flaws in their own reasoning, remains dubious. In this study, we address this que…