Metraipolis › archive › dc_5db7f30d2ede
Diagnosing the Explicit-to-Applied Theory of Mind Gap in LLMs — arXiv:2610.31060
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.31060
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-25T20:53:14.501Z
- SHA-256 of the archived text
69dacc63910cbbab4d81d71f04a5c14e72678c24a6a8db20df389e2898ea33ec- Stable URL of this page
https://metraipolis.com/documento/dc_5db7f30d2ede
How to cite
Diagnosing the Explicit-to-Applied Theory of Mind Gap in LLMs — arXiv:2610.31060 — arXiv (en), https://arxiv.org/abs/2610.31060 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5db7f30d2ede (Metraipolis archive dc_5db7f30d2ede, sha256 69dacc63910cbbab4d81d71f04a5c14e72678c24a6a8db20df389e2898ea33ec)
BibTeX
@misc{dc5db7f30d2ede,
title = {Diagnosing the Explicit-to-Applied Theory of Mind Gap in LLMs — arXiv:2610.31060},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.31060},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5db7f30d2ede; sha256 69dacc63910cbbab4d81d71f04a5c14e72678c24a6a8db20df389e2898ea33ec},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.31060. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
As large language models are deployed in high-stakes roles such as therapy, education, and personal assistance, their social cognition must be assessed reliably. Current Theory of Mind (ToM) benchmarks report aggregate scores that conflate two distinct competencies: explicit ToM (recognizing what someone believes) and applied ToM (using that belief to predict behavior). Building on this distinction, we investigate the gap and its recoverability using four structurally diverse benchmarks. Aggregate reporting obscures uneven competence: on SimpleToM, Llama-3.1-8B scores 51.5 percent overall while recognizing mental states at 99.4 percent and predicting behavior at only 28.9 percent. To assess recovery, we compare structured prompting, matched in-context learning, and our proposed Belief-State Test-Time Training (BSTTT), which adapts lightweight parameters on answered support questions through action reconstruction. Across Qwen2.5-7B, Llama-3.1-8B, and GPT-oss-20B, recovery depends on the model, task, and intervention. On SimpleToM, reminders combined with BSTTT improve Llama's behavior prediction by 40.8 percentage points; reminders account for most of the overall improvement, with BSTTT adding 1.3 to 5.5 points beyond reminders across models. Contextual examples often outperform weight updates, while BSTTT remains stronger in specific model-benchmark settings. Temporal belief tracking remains difficult despite partial recovery. These findings show that aggregate ToM scores can overstate applied competence while hiding performance recoverable through targeted assistance. We recommend decomposed reporting of explicit and applied accuracy alongside intervention comparisons that establish which failures are recoverable and under what conditions. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.31060 — 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 2026-10-25T20:47:56.901Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).