Metraipolis › archive › dc_eefff1454fa7
Boosting Deepresearch and LongContext Ability with Self-Generated Deepresearch Rollouts Traces — arXiv:2609.20844
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.20844
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-10T01:27:02.697Z
- SHA-256 of the archived text
a6b03d416f3adf085ed0e4465847f58712cd22b9bcc6964bbb8015d8bd4c9e97- Stable URL of this page
https://metraipolis.com/documento/dc_eefff1454fa7
How to cite
Boosting Deepresearch and LongContext Ability with Self-Generated Deepresearch Rollouts Traces — arXiv:2609.20844 — arXiv (en), https://arxiv.org/abs/2609.20844 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eefff1454fa7 (Metraipolis archive dc_eefff1454fa7, sha256 a6b03d416f3adf085ed0e4465847f58712cd22b9bcc6964bbb8015d8bd4c9e97)
BibTeX
@misc{dceefff1454fa7,
title = {Boosting Deepresearch and LongContext Ability with Self-Generated Deepresearch Rollouts Traces — arXiv:2609.20844},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.20844},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eefff1454fa7; sha256 a6b03d416f3adf085ed0e4465847f58712cd22b9bcc6964bbb8015d8bd4c9e97},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.20844. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Deepresearch (DR) agents interact with real-world web environments through multi-turn search and visit, causing their contexts to grow rapidly over time. We observe that, even after DR Agentic Reinforcement Learning (DR-RL), 61.6% of the model's remaining prediction errors can still be attributed to insufficient long-context understanding, including longcontext hallucination and failures in cross-document evidence integration. It motivates us to further break the bottleneck of DR-RL by strengthening the model's long-context ability. However, effective LongContext training requires more than simply increasing context length. To bridge the data gap, we propose `DR Rollouts to LongContext-QA (DR-to-Long)'. The method repurposes DR-RL trajectories, which naturally contain search histories, visited webpages, evidence snippets, and final-answer supervision. It then replaces the compact snippets and webpage summaries in each trajectory with the full contents of their corresponding URLs, producing substantially longer multi-document contexts while preserving the original evidence relationships. Building on DR-to-Long, we introduce DLD (DR -> LongQA -> DR)-RL. DLD-RL first performs a short DR-RL stage to collect rollout trajectories, which are then converted into LongQA instances at zero annotation cost. The model is subsequently optimized with LongQA-RL to strengthen LongContext ability, followed by full DR-RL to continue improving its DR capability. Experiments show that DLD-RL outperforms standard DR-RL by 7.3% on three Deepresearch benchmarks and improves performance by 13.5% on three long-context benchmarks. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-21 in cs.CL). Categorie arXiv: Computation and Language (cs.CL). Pagina del lavoro: https://arxiv.org/abs/2609.20844 — 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 2026-10-10T01:00:30.253Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-21?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).