微信内可能无法直接打开本站。请点右上角 ··· → 在浏览器打开，或复制链接。
Causal evidence that language models use confidence to drive behaviour
RSS 官方收录 · 可信分层展示
关键摘要
研究证实大语言模型用置信度驱动回答/拒答行为
- 实验通过增强或抑制置信度信号调控模型决策
- 模型在回答与拒答间的选择受置信度直接因果影响
- 发表于《Nature Machine Intelligence》2026年9月
AI 摘要 · 来源可核验
正文提要
Nature Machine Intelligence, Published online: 07 September 2026; doi:10.1038/s42256-026-01293-x
Kumaran et al. show that large language models making decisions on when to answer a question or abstain from answering can be influenced by boosting or suppressing confidence signals in the model.