研究人员开发了一种名为“基于元认知反馈的强化学习”（RLMF）的新方法，以改进大型语言模型（LLMs）表达其不确定性的方式。该方法利用模型对其性能的自我评估来优化其响应并识别有价值的训练数据，其表现优于标准的活动学习技术。实验表明，RLMF显著增强了“忠实校准”，使表达的不确定性与内在置信度保持一致，并提高了LLMs识别和传达其知识边界的能力。 AI
影响 这项研究可能通过提高大型语言模型表达不确定性和避免自信性幻觉的能力，从而带来更可靠、更值得信赖的大型语言模型。
排序理由 该集群描述了一篇详细介绍改进大型语言模型能力的新方法的最新研究论文。
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
arXiv:2606.32032v1 Announce Type: cross Abstract: Metacognition is a critical component of intelligence that describes the ability to monitor and regulate one's own cognitive processes. Yet LLMs exhibit systemic deficiencies in key metacognitive faculties: they hallucinate with h…
Metacognition is a critical component of intelligence that describes the ability to monitor and regulate one's own cognitive processes. Yet LLMs exhibit systemic deficiencies in key metacognitive faculties: they hallucinate with high confidence, fail to recognize knowledge bounda…
Reinforcement learning with metacognitive feedback and metacognitive data selection improve large language model calibration by enabling accurate self-assessment of performance and uncertainty.