Metraipolis › archive › dc_b95b2efc8dcf

GRADUS: Discovering and Recovering Agent Lessons from the Environment's Latent Syllabus — arXiv:2610.27781

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.27781
License (as archived): CC0-1.0
Archived: 2026-10-21T07:06:50.689Z
SHA-256 of the archived text: 5b33de47889cfa1849c89bf500d5054af9e45d45c4c7356e3f034961f02a9499
Stable URL of this page: https://metraipolis.com/documento/dc_b95b2efc8dcf

How to cite

GRADUS: Discovering and Recovering Agent Lessons from the Environment's Latent Syllabus — arXiv:2610.27781 — arXiv (en), https://arxiv.org/abs/2610.27781 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b95b2efc8dcf (Metraipolis archive dc_b95b2efc8dcf, sha256 5b33de47889cfa1849c89bf500d5054af9e45d45c4c7356e3f034961f02a9499)

BibTeX

@misc{dcb95b2efc8dcf,
  title = {GRADUS: Discovering and Recovering Agent Lessons from the Environment's Latent Syllabus — arXiv:2610.27781},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.27781},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b95b2efc8dcf; sha256 5b33de47889cfa1849c89bf500d5054af9e45d45c4c7356e3f034961f02a9499},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.27781. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Training LLM agents for policy-constrained tool environments, such as an airline backend with booking, refund, and cancellation policies, requires tasks that exercise the environment's rules individually and in combination. Existing synthesis pipelines let a model propose tasks and retain only those it immediately solves, biasing training data toward easier behaviors. We introduce Gradus (Graded Rungs of Ascending Difficulty from an Underlying System), which decouples what a task should teach from the concrete task and solution used to teach it. Without seed tasks, Gradus treats a fixed environment as a latent syllabus, extracting guarded behavioral operators that capture how actions depend on current conditions and how their effects reshape subsequent decisions. It composes these operators into increasingly demanding tasks while using corpus-level feedback to target missing outcome combinations. Crucially, Gradus jointly calibrates tasks and solutions against the environment: execution feedback identifies which side to refine while preserving the intended behavioral requirements. Verification thus becomes a recovery signal rather than a terminal filter, coupling coverage-guided exploration with the preservation of difficult examples. From off-the-shelf backbones, the resulting curriculum lifts a 30B model from 59% to 89% on $\tau^2$-Bench and from 47% to 83% on CAR-Bench with an Opus 4.5 proposer, and to 76% and 62%, respectively, with an open-weight Qwen3-235B proposer. These results show the value of jointly controlling which behaviors are explored and which survive into verified training data.

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/2610.27781 — 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 2026-10-21T07:05:43.445Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).