Metraipolis › archive › dc_0e9bb41dd02b
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs — arXiv:2610.23400
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.23400
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-19T08:43:50.447Z
- SHA-256 of the archived text
6088ccf5d3daa3d015bf980007f85ac598655061bb2c7b225cd99fc0fb26ebcf- Stable URL of this page
https://metraipolis.com/documento/dc_0e9bb41dd02b
How to cite
On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs — arXiv:2610.23400 — arXiv (en), https://arxiv.org/abs/2610.23400 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0e9bb41dd02b (Metraipolis archive dc_0e9bb41dd02b, sha256 6088ccf5d3daa3d015bf980007f85ac598655061bb2c7b225cd99fc0fb26ebcf)
BibTeX
@misc{dc0e9bb41dd02b,
title = {On the Tip of the Tongue: First-Token Cueing as a Behavioral Probe of Recall Failures in LLMs — arXiv:2610.23400},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.23400},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0e9bb41dd02b; sha256 6088ccf5d3daa3d015bf980007f85ac598655061bb2c7b225cd99fc0fb26ebcf},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.23400. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models often answer factual questions incorrectly in closed-book settings, and such errors are usually attributed to missing knowledge. Yet models often encode answers they fail to produce. We show that many closed-book errors are first-token recall failures: the model commits to a wrong first token, but once given the right one it completes the answer and often recognizes it as correct. We study this with First-Token Cueing, a behavioral probe that prefills the response with the first token of the gold answer. Model uncertainty peaks at the first answer token, where it already predicts correctness, and among gold prefixes of any length the first token alone accounts for the largest share of recovered errors. Across 16 open models, this one token corrects between a fifth and a half of the errors on multi-token answers in English QA and improves accuracy in all seven languages of a multilingual benchmark. Most of the errors it corrects are also solved by the model's own sampled reasoning, a criterion for encoded knowledge, and models recognize correct starts they do not produce: they continue a correct start more confidently than an incorrect start of the same answer type and prior probability, yet rarely find it without the gold answer. These results locate a recoverable class of factual errors at the first decoding step, where inference-time methods may gain the most. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2610.23400 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-19T08:39:24.869Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).