Metraipolis › archive › dc_0fb2631ea8fe

Co-Evolving Acting and Forecasting for Language Agents — arXiv:2610.20597

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.20597
License (as archived)
CC0-1.0
Archived
2026-10-15T14:23:57.451Z
SHA-256 of the archived text
bcf34a6a047ad7f110b1b45fea317a1fcab23682694fabceb2eb10211273a119
Stable URL of this page
https://metraipolis.com/documento/dc_0fb2631ea8fe

How to cite

Co-Evolving Acting and Forecasting for Language Agents — arXiv:2610.20597 — arXiv (en), https://arxiv.org/abs/2610.20597 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0fb2631ea8fe (Metraipolis archive dc_0fb2631ea8fe, sha256 bcf34a6a047ad7f110b1b45fea317a1fcab23682694fabceb2eb10211273a119)
BibTeX
@misc{dc0fb2631ea8fe,
  title = {Co-Evolving Acting and Forecasting for Language Agents — arXiv:2610.20597},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.20597},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0fb2631ea8fe; sha256 bcf34a6a047ad7f110b1b45fea317a1fcab23682694fabceb2eb10211273a119},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.20597. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Training language agents is bottlenecked by the high cost of long-horizon rollouts, yet standard agentic reinforcement learning (RL) learns only from a terminal scalar reward, leaving the supervision in intermediate steps largely unused. Prior work exploits this supervision through value targets that estimate per-step success probabilities or observation targets that fit environment response. Both are hard to construct well: (1) value targets come from the same single scalar signal for each step in the trajectory, and (2) observation targets, such as web pages and terminal logs, are long and noisy and reach only one step ahead. We propose \name, which adds a third kind of multi-step action targets. On successful on-policy trajectories, \name trains the policy backbone to generate the action sequences that follows each history. Each action is also supervised when forecast from the preceding histories, forming a temporal ensemble that the shared backbone absorbs during training, and this enables acting and forecasting to co-evolve. The policy still acts one step at a time at inference, with no overhead. Across five benchmarks (ALFWorld, ScienceWorld, WebShop, $\tau^2$-bench, and AppWorld), \name (1) achieves stronger aggregate performance than GRPO and RL baselines with predictive auxiliary targets across simulated environments, while improving overall performance on both realistic applications; (2) delivers up to $1.9\times$ and $2.0\times$ higher sample efficiency than ECHO and GRPO, respectively, on ScienceWorld; and (3) achieves these gains with a single policy backbone, without additional models, separate pretraining, or inference overhead. The codebase is available at this https URL .

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.20597 — 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 2026-10-15T13:53:35.987Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).