Metraipolis › archive › dc_2b15a785efed
Widening Harm Evaluation Coverage with a Judge that Reads the World — arXiv:2611.02624
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2611.02624
- License (as archived)
- CC0-1.0
- Archived
- 2026-11-03T10:32:47.889Z
- SHA-256 of the archived text
48fcef231cc282c0c6a55a47f99a766ba846444508a20e1c265edbbf74799982- Stable URL of this page
https://metraipolis.com/documento/dc_2b15a785efed
How to cite
Widening Harm Evaluation Coverage with a Judge that Reads the World — arXiv:2611.02624 — arXiv (en), https://arxiv.org/abs/2611.02624 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2b15a785efed (Metraipolis archive dc_2b15a785efed, sha256 48fcef231cc282c0c6a55a47f99a766ba846444508a20e1c265edbbf74799982)
BibTeX
@misc{dc2b15a785efed,
title = {Widening Harm Evaluation Coverage with a Judge that Reads the World — arXiv:2611.02624},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2611.02624},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2b15a785efed; sha256 48fcef231cc282c0c6a55a47f99a766ba846444508a20e1c265edbbf74799982},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.02624. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
An agent acting in a stateful environment can leave evidence of harm in two places: the record of what it did and said, or the world it changed. Safety evaluations for discovering harm examine one or the other in limited ways, using either an LLM judge over the trajectory or deterministic checks over the final state. Although LLM-based verifiers that inspect the environment exist, they have been used to verify task completion, not to detect and confirm harm in a safety setting. We address this gap with an LLM that has read-only access to the environment state and is given a harm taxonomy. We evaluate this approach in three environments: two from DecodingTrust-Arena and one from OSWorld. We hold the judge model and harm taxonomy fixed, varying only what the judge observes: the trajectory, read-only tools for inspecting the environment, or both channels together. In the two DecodingTrust-Arena environments, combining the trajectory and state-only judges recalls 94 percent of confirmed harms, compared with 53 percent for the trajectory judge alone. On OS-Harm, the union recalls 77 percent, compared with 67 percent. A single judge given both channels does not perform better and produces more false negatives than the state-only judge. We also record disagreements and find that the two channels miss different episodes. These results suggest that combining verdicts from two channel-specific judges is a robust and flexible approach to detecting and verifying harm in agentic settings. 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/2611.02624 — 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 2026-11-03T10:29:48.572Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).