Metraipolis › archive › dc_0779787afa35

Residual Visual Credit Optimization: Conserved Evidence Routing for Multimodal Reinforcement Learning — arXiv:2610.04918

Source
arXiv
Original
https://arxiv.org/abs/2610.04918
License (as archived)
CC0-1.0
Archived
2026-10-06T05:28:48.218Z
SHA-256 of the archived text
56719b3869952f698e5706305d4f273a4b6009de1c311e8378215c5a8d2b6c28
Stable URL of this page
https://metraipolis.com/documento/dc_0779787afa35

How to cite

Residual Visual Credit Optimization: Conserved Evidence Routing for Multimodal Reinforcement Learning — arXiv:2610.04918 — arXiv, https://arxiv.org/abs/2610.04918 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_0779787afa35 (Metraipolis archive dc_0779787afa35, sha256 56719b3869952f698e5706305d4f273a4b6009de1c311e8378215c5a8d2b6c28)
BibTeX
@misc{dc0779787afa35,
  title = {Residual Visual Credit Optimization: Conserved Evidence Routing for Multimodal Reinforcement Learning — arXiv:2610.04918},
  howpublished = {arXiv},
  url = {https://arxiv.org/abs/2610.04918},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_0779787afa35; sha256 56719b3869952f698e5706305d4f273a4b6009de1c311e8378215c5a8d2b6c28},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv, https://arxiv.org/abs/2610.04918. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning with verifiable rewards scales multimodal reasoning, but an outcome reward says how much a trajectory is worth, not how that value should be spread over the decisions that produced it. We introduce Residual Visual Credit Optimization (RVCO), which treats token credit as a conserved routing problem. A controlled visual intervention yields a per-token evidence response; robust within-trajectory coordinates remove incidental scale; and a budgeted entropic router distributes a fixed amount of sequence utility according to perceptual dependence. A residual support path guarantees positive credit at every valid position, and an analytic correction restores the prescribed credit mass exactly. The resulting field is selective, bounded, full-support, and invariant to response-local score shifts, and recovers hard token selection as a limiting case. Across four model families and seven reasoning benchmarks, RVCO improves accuracy over strong RLVR baselines while maintaining late-stage optimization stability, corruption robustness, and competitive training cost. Rewards, rollouts, and the group-relative advantage estimator are unchanged; only the geometry of token-level credit differs.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.04918 — 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 2026-10-06T05:24:20.427Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).