Metraipolis › archive › dc_2e47d7658749

JET: Judge-Guided Evolution at Test Time for Agent Programs — arXiv:2609.34126

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34126
License (as archived)
CC0-1.0
Archived
2026-10-07T21:02:10.121Z
SHA-256 of the archived text
c1d127f037f768ed255dd466386bac3207ed2a454d5c64c03e8aeb67bcc1b4b9
Stable URL of this page
https://metraipolis.com/documento/dc_2e47d7658749

How to cite

JET: Judge-Guided Evolution at Test Time for Agent Programs — arXiv:2609.34126 — arXiv (en), https://arxiv.org/abs/2609.34126 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2e47d7658749 (Metraipolis archive dc_2e47d7658749, sha256 c1d127f037f768ed255dd466386bac3207ed2a454d5c64c03e8aeb67bcc1b4b9)
BibTeX
@misc{dc2e47d7658749,
  title = {JET: Judge-Guided Evolution at Test Time for Agent Programs — arXiv:2609.34126},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34126},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2e47d7658749; sha256 c1d127f037f768ed255dd466386bac3207ed2a454d5c64c03e8aeb67bcc1b4b9},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34126. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

An agent's executable program governs how it uses tools, processes observations, and responds to failures. Evolving this program at test time can help adaptation, but deciding which changes to retain is difficult when true rewards are unavailable. Execution traces provide evidence of agent behavior, yet interpreting that evidence requires a judge that remains useful as tasks and candidate programs change. We introduce Judge-Guided Evolution at Test Time (JET), which evolves an executable judge on labeled source trajectories, then freezes and transfers it to guide target-side program evolution. The judge supplies scores and diagnostic feedback without target evaluator access or model-weight updates. On unseen WebShop tasks, JET achieves approximately 13% higher mean reward than fixed-rubric guidance when evolution begins from an unevolved program (cold start) and 4% higher when it begins from one already optimized on source tasks (warm start), with a 36% relative improvement in cold-start exact success. An exact-judge control on PushT, where the judge reconstructs the scoring rule from observations, shows that without judge error, program search becomes the bottleneck. Analyses identify useful reward-prediction logic in the evolved code and show that better final selection alone cannot explain the gains. These results support executable judge transfer for program adaptation under evaluator-preserving task shifts.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Software Engineering (cs.SE) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34126 — 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 2026-10-07T20:29:02.869Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).