Metraipolis › archive › dc_80490bf3d166

TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing — arXiv:2610.41143

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.41143
License (as archived): CC0-1.0
Archived: 2026-11-02T07:49:25.917Z
SHA-256 of the archived text: 992af7e7f8119960736755c1d885d1ab673d0922c37aa0ea1b1d93cc3af47e90
Stable URL of this page: https://metraipolis.com/documento/dc_80490bf3d166

How to cite

TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing — arXiv:2610.41143 — arXiv (en), https://arxiv.org/abs/2610.41143 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_80490bf3d166 (Metraipolis archive dc_80490bf3d166, sha256 992af7e7f8119960736755c1d885d1ab673d0922c37aa0ea1b1d93cc3af47e90)

BibTeX

@misc{dc80490bf3d166,
  title = {TalentRL: Strategy-Guided Reinforcement Learning for Hierarchical Skill Routing — arXiv:2610.41143},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.41143},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_80490bf3d166; sha256 992af7e7f8119960736755c1d885d1ab673d0922c37aa0ea1b1d93cc3af47e90},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.41143. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Natural-language skills enable large language model (LLM) agents to reuse procedures across tasks, yet most systems retrieve individual skills from a flat SkillBank without explicitly modeling the high-level strategy that governs their selection and use. We introduce TalentRL, a strategy-guided reinforcement learning framework for hierarchical routing over procedural memory. Coarse-grained talents represent recurring strategies for skill use, while fine-grained skills provide executable procedures. For each task, the policy model selects a talent that guides execution and softly prioritizes its associated skills during similarity-based retrieval. However, since task outcomes depend jointly on talent selection, retrieved skill context, and downstream execution, terminal rewards alone cannot isolate the contribution of the routing decision. TalentRL therefore trains talent routing with reward contrasts between rollouts of the same task under matched interventions on the talent-conditioned context, and optimizes downstream execution with group-relative advantages. Across seven open-domain and multi-hop question-answering benchmarks, talent-aware supervised fine-tuning improves average success rate by 7.53%, 6.71%, and 9.21% over the corresponding zero-shot baselines for Qwen2.5-7B, Qwen3-8B, and Llama-3.1-8B, respectively. Reinforcement learning further improves these SFT checkpoints by 7.58% to 9.32%, and TalentRL achieves the highest average success rate among all compared methods on all three backbones. On the agentic benchmarks ALFWorld and WebShop, TalentRL achieves the highest ALFWorld success rate (99.3%) and the highest WebShop score (90.7) among all compared methods.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.LG, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2610.41143 — 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 2026-11-02T07:40:00.174Z dalla pagina https://arxiv.org/list/cs.LG/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).