Metraipolis › archive › dc_9bee7749c5bb
Before the Next Rollout: Diagnosing Skill Utility Shifts from Policy Updates — arXiv:2610.39358
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.39358
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-31T06:09:16.402Z
- SHA-256 of the archived text
413208d7dbea3c9b7a9cd332b9ce1edf5b13c8bb498147b2fcc2efa4cb625d23- Stable URL of this page
https://metraipolis.com/documento/dc_9bee7749c5bb
How to cite
Before the Next Rollout: Diagnosing Skill Utility Shifts from Policy Updates — arXiv:2610.39358 — arXiv (en), https://arxiv.org/abs/2610.39358 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_9bee7749c5bb (Metraipolis archive dc_9bee7749c5bb, sha256 413208d7dbea3c9b7a9cd332b9ce1edf5b13c8bb498147b2fcc2efa4cb625d23)
BibTeX
@misc{dc9bee7749c5bb,
title = {Before the Next Rollout: Diagnosing Skill Utility Shifts from Policy Updates — arXiv:2610.39358},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.39358},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_9bee7749c5bb; sha256 413208d7dbea3c9b7a9cd332b9ce1edf5b13c8bb498147b2fcc2efa4cb625d23},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.39358. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Skill-evolution methods typically analyze execution trajectories to guide agent skill refinement. However, policy updates can change the behavioral effects and marginal utility of an unchanged skill. Attribution based on earlier trajectories may therefore no longer reflect the skill's contribution under the updated policy. Moreover, trajectory failures can arise from multiple factors, making individual skill contributions difficult to isolate. We evaluate whether policy-update signals contain predictive information about changes in skill marginal utility. At fixed contexts from pre-update training rollouts, we measure how policy updates change the contrast in token preferences between skill and control conditions. Combining these interaction shifts with advantage estimates yields reward-directed readouts for diagnosing utility changes. The readouts reuse existing training rollouts and require only forward passes for single-step action scoring, without additional post-update rollout evaluation. Our experiments show that policy updates can alter skill marginal utility and that policy-update readouts carry diagnostic information for identifying skills with declining utility. Integrating readout-guided prioritization into skill evolution improves downstream performance while reducing policy and editor token consumption. 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/2610.39358 — 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 2026-10-31T06:07:08.517Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).