Repairing the Know-Say Gap: A No-Finetuning Probe-to-Logit Confidence Controller

Abstract Ask a language model how confident it is and the answer is usually uninformative, even though the model holds a usable signal about its own correctness in its hidden states. We argue that this know-say gap is better explained as a routing bottleneck than as a missing capability, and show that it can be repaired without changing a single weight. A linear probe on a mid-layer state, paired with a ten-weight projection onto the confidence-token logits and fit on a few hundred labelled examples, makes the model verbalise the calibrated confidence it otherwise withholds. Across five base models on TriviaQA this controller out-discriminates five-sample self-consistency on four of five models at a fifth of the inference cost, matches or beats P(True) at every model below 72B, and is the only single-pass method whose confidence never inverts. In a controlled installation study, the head-to-logit route reaches verbal AUROC2 0.765 and survives a subsequent alignment pass, while fine-tuning the model to emit confidence leaves it near chance (0.57-0.58). Under a criterion-validated psychometric screen, the controller is the only signal that stays valid across models and binarisations while every baseline fails somewhere.