Metraipolis › archive › dc_4ae83c8d48b2
CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty — arXiv:2610.22292
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.22292
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T10:04:25.574Z
- SHA-256 of the archived text
ad9841730886cf335814f252ee34773de62e4f19cba9e3e28efafd1a0f10f11f- Stable URL of this page
https://metraipolis.com/documento/dc_4ae83c8d48b2
How to cite
CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty — arXiv:2610.22292 — arXiv (en), https://arxiv.org/abs/2610.22292 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4ae83c8d48b2 (Metraipolis archive dc_4ae83c8d48b2, sha256 ad9841730886cf335814f252ee34773de62e4f19cba9e3e28efafd1a0f10f11f)
BibTeX
@misc{dc4ae83c8d48b2,
title = {CAT-alyzing Honesty: Continuous Adversarial Training for Transferable LLM Honesty — arXiv:2610.22292},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.22292},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4ae83c8d48b2; sha256 ad9841730886cf335814f252ee34773de62e4f19cba9e3e28efafd1a0f10f11f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22292. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language models (LLMs) are becoming prevalent in scenarios where honesty is critical. However, despite alignment training, these models still exhibit dishonest behaviour, such as outputting claims that contradict some knowledge or are not supported by what is accessible in context. We introduce Continuous Adversarial Training for Consistent Honesty (CATCH), inspired by a method originally developed to robustly refuse harmful queries. CATCH trains the model towards an honest answer and away from a dishonest one while maintaining the model's capabilities. We evaluate two architectures (Llama-3.3-70B-Instruct and Qwen3-3.6-27B). We show that trained with only hundreds of targets built from self-reporting scenarios, CATCH learns an honest behaviour that robustly generalizes and transfers across distinct benchmarks. It consistently reduces dishonesty under pressure on MASK, increases confessions in the multi-agent game of Among Us, and improves the discovery rates of the model's hidden behaviours in AuditBench. We show that CATCH outperforms prompting and supervised fine-tuning in improving honesty, and it is on par with state-of-the-art honesty activation steering. However, CATCH surpasses honesty steering in maintaining utility, thanks to a novel choice of utility data. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL); Computers and Society (cs.CY). Pagina del lavoro: https://arxiv.org/abs/2610.22292 — 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 2026-10-16T08:44:47.777Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).