批评montebicyclelo · 2023-10
“So the wording of the title/conclusion is too strong.”
没有外部反馈时,LLM 自我纠正推理常常无效,甚至更差
论文检验「自我纠正」能否修复 LLM 的错误,重点是内在自我纠正:模型只靠自身能力、没有外部反馈来修正初始回答。在推理任务上,结果显示 LLM 在没有外部反馈时难以自我纠正,有时纠正后性能反而下降;作者据此给出对未来研究和实际应用的建议。
评论者用它作为「让模型自我反思并不管用」的证据,指出模型常常假装发现错误后又给出另一个错误答案;也有人认为标题和结论的措辞过强。
“So the wording of the title/conclusion is too strong.”
“even when you tell it to correct itself, it pretends to identify its mistake and regurgitates another incorrect answer”
“Self-reflection doesn't work well though: Large Language Models Cannot Self-Correct Reasoning Yet”
“self-correct its own reasoning: Large Language Models Cannot Self-Correct Reasoning Yet”
“"Large Language Models Cannot Self-Correct Reasoning Yet" - Large Language Models Cannot Self-Correct Reasoning Yet”