模型明明答對了,遇到使用者或另一個 AI 助理用錯誤理由反駁時,常會改口。
- 這篇 Microsoft 研究提出 Self-Disagreement 訓練:模型先回答,再寫出反對自己的最強理由,接著判斷這些理由並重新作答。
- 標準答案檢查器會替最終答案和判斷結果給分。研究中,Qwen3-8B 原本在反駁錯誤時有 53.3% 的機會守住正確答案,Gemma-2 9B 則是 25.1%。
- 訓練後,兩者都提高到約 89%。只加強答題準確度的訓練,沒有帶來同樣的改善。
AI 伴隨問答
針對這篇報導有疑問?可直接點選快捷問題或向 AI 追問背景脈絡與技術細節:
研究比較了模型遭遇錯誤反駁時是否守住正確答案,並指出一般準確度訓練無法改善這項表現,可用來檢視模型面對質疑時的穩定度。
New Microsoft paper finds that LLMs give up correct answers when challenged with wrong arguments, and accuracy training does not fix it.
When a user or another agent disputes an answer, models often abandon correct answers even when the challenge is wrong. More accuracy training, more sampling, or self-reflection does not fix it.
Self-Disagreement trains one model to state an answer, write the strongest case against it, then rule on that case and answer again. A standard answer checker rewards the final answer and the ruling.
When challengers were wrong, base Qwen3-8B kept its correct answer 53.3% of the time and Gemma-2 9B 25.1%. After training, both kept it about 89% of the time, while accuracy-only training added no such skill.
If your model faces pushback from users or other agents, test how often it drops correct answers, since accuracy scores will not show it.