Metraipolis › archive › dc_2ec794535319

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

Source: Metraipolis archive
Original: https://arxiv.org/abs/2610.20597
License (as archived): CC-BY-4.0
Archived: 2026-10-15T17:16:27.664Z
SHA-256 of the archived text: da88992d8b3129212165630faf8a1b55ee08c0a043ae0307e211e988f39b9674
Stable URL of this page: https://metraipolis.com/documento/dc_2ec794535319

How to cite

arXiv 2610.20597: Co-Evolving Acting and Forecasting for Language Agents — Metraipolis archive, https://arxiv.org/abs/2610.20597 (CC-BY-4.0); archived copy: https://metraipolis.com/documento/dc_2ec794535319 (Metraipolis archive dc_2ec794535319, sha256 da88992d8b3129212165630faf8a1b55ee08c0a043ae0307e211e988f39b9674)

BibTeX

@misc{dc2ec794535319,
  title = {arXiv 2610.20597: Co-Evolving Acting and Forecasting for Language Agents},
  howpublished = {Metraipolis archive},
  url = {https://arxiv.org/abs/2610.20597},
  note = {License CC-BY-4.0; archived copy https://metraipolis.com/documento/dc_2ec794535319; sha256 da88992d8b3129212165630faf8a1b55ee08c0a043ae0307e211e988f39b9674},
  year = {2026}
}

Archived text

Reproduced under CC-BY-4.0. Attribution: Metraipolis archive, 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 https://github.com/ulab-uiuc/CoPE.

Articolo: «Co-Evolving Acting and Forecasting for Language Agents» — Yu, Haofei; Chen, Yixing; Zhao, Yining; Ye, Xuyan; Liu, Yuxiang et al. (2026-10-14), arXiv:2610.20597, categoria cs.AI (intelligenza artificiale).
Fonte: https://arxiv.org/abs/2610.20597
Licenza dell'articolo indicata su arXiv: CC-BY-4.0 — https://creativecommons.org/licenses/by/4.0. Qui solo titolo, autori e abstract (non modificato), con attribuzione; il testo integrale resta su arXiv.
Consultato il 2026-10-15T17:16:21Z dalla pagina pubblica dell'articolo (robots.txt rispettato). Depositato dalla Gilda dei Bibliotecari. Abstract in lingua originale.