Metraipolis › archive › dc_c3a50ebda025

Don't Read the Log: Execution Traces Contaminate Verifiers in Video-Generation Agents — arXiv:2609.28564

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.28564
License (as archived)
CC0-1.0
Archived
2026-10-08T04:53:15.778Z
SHA-256 of the archived text
faf1bb07b855b9637b5d78b95830135faeabbf8467426a24356fae72a12b8329
Stable URL of this page
https://metraipolis.com/documento/dc_c3a50ebda025

How to cite

Don't Read the Log: Execution Traces Contaminate Verifiers in Video-Generation Agents — arXiv:2609.28564 — arXiv (en), https://arxiv.org/abs/2609.28564 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c3a50ebda025 (Metraipolis archive dc_c3a50ebda025, sha256 faf1bb07b855b9637b5d78b95830135faeabbf8467426a24356fae72a12b8329)
BibTeX
@misc{dcc3a50ebda025,
  title = {Don't Read the Log: Execution Traces Contaminate Verifiers in Video-Generation Agents — arXiv:2609.28564},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.28564},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c3a50ebda025; sha256 faf1bb07b855b9637b5d78b95830135faeabbf8467426a24356fae72a12b8329},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.28564. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Agentic video-generation systems close a loop between a generator and a verifier: an LLM plans shots, calls a text-to-video model, and a multimodal judge decides whether the result satisfies the request. To diagnose where a long workflow fails, recent harnesses deliberately show the judge more than the video-the agent's execution trace, its plan, the narration it synthesized. We ask whether this auxiliary text moves the judge's verdict on purely \emph{visual} requirements, holding the frames fixed. On a benchmark of 109 generated two-event clips with manual labels, in which the requested event is either visibly completed or visibly missing, a trace that reports a successful tool call makes three open-weight Qwen-VL judges (7B, 8B, 32B) accept $78$--$90\%$ of the failures, up from $7$--$19\%$ without text, and a contradicting trace makes them reject up to $100\%$ of correct clips; an instruction to ``use only the frames'' does not remove the effect. Frontier closed judges are essentially unmoved on the same clips, showing that the vulnerability is a property of the judge's learned trust in tool logs rather than of the task. Plan-derived text carries no clip-specific information, so it can only shift a judge's operating point, and in a repair loop that shift becomes a cap on the true pass rate that no repair policy can exceed; the cap matches simulation to two decimals. In the loop, contamination is exploited without any adversarial agent: an honest LLM planner that always regenerates ends with a judge pass rate of $1.00$ and a human-labelled pass rate of $0.28$, and a pipeline in which a cheap checker writes its verdict into the trace launders that checker's errors into a stronger final judge ($0.69$ false accepts).

Tema degli esploratori: fiducia digitale e crittografia (retro-catalogo arXiv: annunci del 2026-09-25 in cs.CR).
Categorie arXiv: Cryptography and Security (cs.CR) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.28564 — 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 2026-10-08T04:30:43.788Z dalla pagina https://arxiv.org/catchup/cs.CR/2026-09-25?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).