Causal Tongue-Tie: LLMs Can Encode Causal Direction, But Their Yes/No Outputs Fail to Express 事件

Causal Tongue-Tie: LLMs Can Encode Causal Direction, But Their Yes/No Outputs Fail to Express arXiv:2605.25891v1 Announce Type: new Abstract: We find a mismatch between what large language models encode about a causal question and what they answer. On anti-commonsense CLadder items, a fixed linear probe recovers the evidence-supported answer from the model's hidden state (accuracy approximately 0.97), while the spoken Yes/No reverts to the commonsense one (accuracy approximately 0.5). We call t