Metraipolis › archive › dc_69797c8d1148
Improving scalable oversight with co-trained monitors — arXiv:2609.36049
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.36049
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:25:11.109Z
- SHA-256 of the archived text
b38f3bf9e9e9f8f1e5b66e9342087917bc2b09e45e109b1104533d8202cc14b2- Stable URL of this page
https://metraipolis.com/documento/dc_69797c8d1148
How to cite
Improving scalable oversight with co-trained monitors — arXiv:2609.36049 — arXiv (en), https://arxiv.org/abs/2609.36049 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_69797c8d1148 (Metraipolis archive dc_69797c8d1148, sha256 b38f3bf9e9e9f8f1e5b66e9342087917bc2b09e45e109b1104533d8202cc14b2)
BibTeX
@misc{dc69797c8d1148,
title = {Improving scalable oversight with co-trained monitors — arXiv:2609.36049},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.36049},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_69797c8d1148; sha256 b38f3bf9e9e9f8f1e5b66e9342087917bc2b09e45e109b1104533d8202cc14b2},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36049. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Worker-monitor setups are a promising approach to AI oversight, but training workers against fixed monitors can incentivize monitor evasion. We study whether this failure mode can be avoided by co-training the monitor alongside the worker, and explore both supervised and self-supervised approaches. In the supervised setting, we prove a characterization: monitoring is possible with vanishing error and query rates exactly when the class of possible monitor functions has finite Littlestone dimension. This connects worker monitoring with an established literature on adversarial online learning. For self-supervision, we propose a co-training procedure based on test-time distillation: the monitor uses additional test-time compute to generate training labels, then trains its standard-compute policy on those labels. For majority-vote labels, we give a finite-sample sharpening guarantee under adaptive worker distributions with action coverage, that shows that the monitor's verdicts converge to its initial modal verdicts. We stress-test the former in code-security settings where the worker is trained adversarially to fool the monitor. Our results suggest that adaptive monitors are better at keeping pace with evolving worker strategies, while fixed monitors are more vulnerable to evasion. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI). Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.36049 — 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 2026-10-07T15:54:53.760Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).