Metraipolis › archive › dc_b4c01fa68ddd
ForkPilot: Self-Evolving Policy for Retrospective Search in Long-Horizon Agents — arXiv:2610.04889
- Source
- arXiv
- Original
- https://arxiv.org/abs/2610.04889
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-06T05:24:52.265Z
- SHA-256 of the archived text
6807c760ecc45f4722d4480ca3cf4d054b25ddbb589a176104b7ee8d77747eac- Stable URL of this page
https://metraipolis.com/documento/dc_b4c01fa68ddd
How to cite
ForkPilot: Self-Evolving Policy for Retrospective Search in Long-Horizon Agents — arXiv:2610.04889 — arXiv, https://arxiv.org/abs/2610.04889 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_b4c01fa68ddd (Metraipolis archive dc_b4c01fa68ddd, sha256 6807c760ecc45f4722d4480ca3cf4d054b25ddbb589a176104b7ee8d77747eac)
BibTeX
@misc{dcb4c01fa68ddd,
title = {ForkPilot: Self-Evolving Policy for Retrospective Search in Long-Horizon Agents — arXiv:2610.04889},
howpublished = {arXiv},
url = {https://arxiv.org/abs/2610.04889},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_b4c01fa68ddd; sha256 6807c760ecc45f4722d4480ca3cf4d054b25ddbb589a176104b7ee8d77747eac},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.04889. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Interactive language-model agents increasingly solve complex tasks through long-horizon, multi-call reasoning, where errors in beliefs or actions can compound across tool interactions. Retrospective search can recover from such failures but is prone to misallocation. Delayed outcomes obscure the contribution of intermediate search decisions, leading to Attribution Complexity, while evolving execution evidence leads to Adaptation Complexity, where previously learned estimates become stale. To address these challenges, we first introduce Search Value Dynamics (SVD), which characterizes the evolving trade-off between the gain and cost of retrospective search. Building on SVD, we propose ForkPilot, a self-evolving two-stage policy-learning framework. In the first stage, ForkPilot learns a search-value policy offline from completed trajectories through automatically constructed outcome comparisons. In the second stage, it makes search decisions based on current observations and then self-evolves by incorporating newly completed trajectories into subsequent policy updates. We evaluate ForkPilot across 6 diverse benchmarks and 7 widely used LLM backbone families, including four open-source families, GPT-5.6 Sol, and Opus 4.8 in a production agentic system, against 9 competitive baselines, including a real-world harness deployment used by hundreds of thousands of paid users. ForkPilot achieves comparable state-of-the-art performance while reducing token usage by up to 59.2%, demonstrating its efficacy. 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/2610.04889 — 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 2026-10-06T05:24:05.529Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).