Metraipolis › archive › dc_137352c6ba16
MMSearch-OPD: Easy Made Hard, Hard Made Learnable for Long-Horizon Multimodal Search Agents — arXiv:2610.22240
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.22240
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T10:36:48.753Z
- SHA-256 of the archived text
ad2cfd3048f36a2640aff732eab891656b8651ac5c9fda6b0f3716e489aeff29- Stable URL of this page
https://metraipolis.com/documento/dc_137352c6ba16
How to cite
MMSearch-OPD: Easy Made Hard, Hard Made Learnable for Long-Horizon Multimodal Search Agents — arXiv:2610.22240 — arXiv (en), https://arxiv.org/abs/2610.22240 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_137352c6ba16 (Metraipolis archive dc_137352c6ba16, sha256 ad2cfd3048f36a2640aff732eab891656b8651ac5c9fda6b0f3716e489aeff29)
BibTeX
@misc{dc137352c6ba16,
title = {MMSearch-OPD: Easy Made Hard, Hard Made Learnable for Long-Horizon Multimodal Search Agents — arXiv:2610.22240},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.22240},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_137352c6ba16; sha256 ad2cfd3048f36a2640aff732eab891656b8651ac5c9fda6b0f3716e489aeff29},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.22240. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Multimodal deep search is moving from short queries to long-horizon investigations that continually acquire evidence. However, training such agents remains constrained by limited challenging training questions and insufficient process feedback. We introduce MMSearch-OPD, combining controllable data refinement with joint reinforcement learning (RL) and on-policy distillation (OPD). Our Refine pipeline performs targeted expansion of reasoning chains derived from existing simple multimodal questions. Parameterized transformations control search depth and cross-modal evidence requirements while reusing existing image--question pairs. For training, answer rewards and lightweight behavior penalties guide RL, while the reward gap modulates token-level teacher guidance on imperfect student trajectories. An exponentially decaying distillation coefficient reduces teacher dependence as training proceeds. Experiments show gains from data refinement and further improvements from joint training after SFT, indicating stronger long-horizon multimodal deep search capabilities. Tema degli esploratori: statistica e scienza dei dati (elenco arXiv cs.CV, nuovi lavori del giorno). Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV). Pagina del lavoro: https://arxiv.org/abs/2610.22240 — 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 2026-10-16T08:52:02.726Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).