Metraipolis › archive › dc_f4b83a6ea043

Evading Chain-of-Thought Monitoring Through Model Poisoning — arXiv:2608.02820

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.02820
License (as archived)
CC0-1.0
Archived
2026-10-31T12:18:15.473Z
SHA-256 of the archived text
2b2513bb0ba019d5ee04301f0cd2f834d84b3c3aa7ad1b5f7c26f84d8bf7813e
Stable URL of this page
https://metraipolis.com/documento/dc_f4b83a6ea043

How to cite

Evading Chain-of-Thought Monitoring Through Model Poisoning — arXiv:2608.02820 — arXiv (en), https://arxiv.org/abs/2608.02820 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f4b83a6ea043 (Metraipolis archive dc_f4b83a6ea043, sha256 2b2513bb0ba019d5ee04301f0cd2f834d84b3c3aa7ad1b5f7c26f84d8bf7813e)
BibTeX
@misc{dcf4b83a6ea043,
  title = {Evading Chain-of-Thought Monitoring Through Model Poisoning — arXiv:2608.02820},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.02820},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f4b83a6ea043; sha256 2b2513bb0ba019d5ee04301f0cd2f834d84b3c3aa7ad1b5f7c26f84d8bf7813e},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.02820. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Chain-of-thought (CoT) monitoring is an increasingly important component of AI safety stacks but relies on the assumption that a model's reasoning trace is informative about its actions. This work studies the limits of CoT monitoring through the lens of model poisoning. We demonstrate that backdoors can be implanted into reasoning models to elicit an attacker-chosen behavior while their CoT traces appear entirely benign. We find that these CoT-Hidden backdoors can be induced through simple fine-tuning recipes across reasoning-model architectures and sizes. When direct poisoning is ineffective, we introduce a curriculum training approach that progressively teaches the model to produce an attacker-chosen output while concealing the behavior from its reasoning traces. These findings suggest that CoT monitoring may be better framed as a question about the consistency between a model's reasoning trace and its final response than as anomaly detection within a trace. We further examine the mechanisms that allow models to suppress evidence of the target behavior from their reasoning traces. Causal interventions locate a trigger-conditioned activation pathway that does not depend on the visible reasoning, and residual stream verbalizations provide an anomaly warning near answer generation, but do not identify the trigger, target, or backdoor mechanism.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-05 in cs.AI).
Categorie arXiv: Cryptography and Security (cs.CR) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2608.02820 — 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 2026-10-31T12:05:52.955Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-05?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).