Metraipolis › archive › dc_2fa0a928d3ec

F$^{2}$DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch Workflows — arXiv:2609.19827

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.19827
License (as archived)
CC0-1.0
Archived
2026-10-10T03:48:07.807Z
SHA-256 of the archived text
1f16f0489de892ea57e11852e943acae078874c72cfec0341461f70dcba6dbdb
Stable URL of this page
https://metraipolis.com/documento/dc_2fa0a928d3ec

How to cite

F$^{2}$DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch Workflows — arXiv:2609.19827 — arXiv (en), https://arxiv.org/abs/2609.19827 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_2fa0a928d3ec (Metraipolis archive dc_2fa0a928d3ec, sha256 1f16f0489de892ea57e11852e943acae078874c72cfec0341461f70dcba6dbdb)
BibTeX
@misc{dc2fa0a928d3ec,
  title = {F$^2$DR: A Fine-Grained Full-Pipeline Reward Framework for DeepSearch Workflows — arXiv:2609.19827},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.19827},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_2fa0a928d3ec; sha256 1f16f0489de892ea57e11852e943acae078874c72cfec0341461f70dcba6dbdb},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.19827. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

With the widespread industrial deployment of Large Language Models (LLMs), DeepSearch has emerged as the dominant paradigm for resolving complex user queries. It typically operates through an iterative closed-loop workflow consisting of planning and reflection, information retrieval, and answer generation. However, existing reward models (RMs) and evaluation benchmarks are primarily designed for static single-turn tasks, failing to capture the full-pipeline complexity of DeepSearch workflows. To address this limitation, we propose F2DR, a fine-grained full-pipeline DeepSearch reward framework. F2DR evaluates DeepSearch workflows across three dimensions: Content, Trajectory, and Answer, enabling comprehensive process-level assessment. We further construct DeepSearch RM-Bench, a dedicated benchmark for evaluating RMs in DeepSearch scenarios. Extensive experiments demonstrate that F2DR achieves significantly higher evaluation consistency than self-evaluation-based baselines, while DeepSearch RM-Bench exhibits strong discriminative capability across existing open-source RMs. We will publicly release the complete DeepSearch RM-Bench dataset soon.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-18 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.19827 — 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 2026-10-10T03:24:08.134Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-18?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).