[2603.17839] How do LLMs Compute Verbal Confidence

Computer Science > Computation and Language arXiv:2603.17839 (cs) [Submitted on 18 Mar 2026 (v1), last revised 31 Mar 2026 (this version, v2)] Title:How do LLMs Compute Verbal Confidence Authors:Dharshan Kumaran, Arthur Conmy, Federico Barbero, Simon Osindero, Viorica Patraucean, Petar Velickovic View a PDF of the paper titled How do LLMs Compute Verbal Confidence, by Dharshan Kumaran and 5 other authors View PDF HTML (experimental) Abstract:Verbal confidence -- prompting LLMs to state their confidence as a number or category -- is widely used to extract uncertainty estimates from black-box models. However, how LLMs internally generate such scores remains unknown. We address two questions: first, when confidence is computed - just-in-time when requested, or automatically during answer generation and cached for later retrieval; and second, what verbal confidence represents - token log-probabilities, or a richer evaluation of answer quality? Focusing on Gemma 3 27B and Qwen 2.5 7B, we provide convergent evidence for cached retrieval. Activation steering, patching, noising, and swap experiments reveal that confidence representations emerge at answer-adjacent positions before appearing at the verbalization site. Attention blocking pinpoints the information flow: confidence is gathered from answer tokens, cached at the first post-answer position, then retrieved for output. Critically, linear probing and variance partitioning reveal that these cached representations explain subs...