Metraipolis › archive › dc_ddefe2d34dca

Perception or Reasoning? Diagnosing and Tackling Reinforcement Learning Bottlenecks in VLMs — arXiv:2610.21380

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.21380
License (as archived): CC0-1.0
Archived: 2026-10-16T10:35:18.262Z
SHA-256 of the archived text: 6da1f82ff8d1ef1fbabe788181e066f016d783dae808c15ca0f3b0b2a6984735
Stable URL of this page: https://metraipolis.com/documento/dc_ddefe2d34dca

How to cite

Perception or Reasoning? Diagnosing and Tackling Reinforcement Learning Bottlenecks in VLMs — arXiv:2610.21380 — arXiv (en), https://arxiv.org/abs/2610.21380 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ddefe2d34dca (Metraipolis archive dc_ddefe2d34dca, sha256 6da1f82ff8d1ef1fbabe788181e066f016d783dae808c15ca0f3b0b2a6984735)

BibTeX

@misc{dcddefe2d34dca,
  title = {Perception or Reasoning? Diagnosing and Tackling Reinforcement Learning Bottlenecks in VLMs — arXiv:2610.21380},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21380},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ddefe2d34dca; sha256 6da1f82ff8d1ef1fbabe788181e066f016d783dae808c15ca0f3b0b2a6984735},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21380. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning has brought strong gains to language model reasoning, but its improvements on visual tasks remain less consistent. One difficulty is that visual task failures can arise from different sources: the same benchmark may be limited by perception for one model and by reasoning for another. We introduce Perception and Reasoning Attribution (PARA), which measures the remaining perceptual and reasoning headroom of each model--benchmark pair by comparing how many current errors are recovered with more accurate visual information versus a stronger reasoner. Across 4 vision-language models and 17 multimodal benchmarks, PARA broadly agrees with the expected characteristics of existing benchmarks while revealing substantial model-specific variation. More importantly, it shows that RL gains become smaller as perceptual headroom increases, but larger when reasoning explains more of the recoverable errors, suggesting that current RL exploits reasoning headroom more effectively than perceptual headroom. Motivated by this gap, we analyze how existing RL objectives affect visual perception and find that reward- and token-level signals influence visual attention only indirectly. We therefore propose GRAD, a group-relative attention distillation objective that uses the best rollout in each group to directly guide the visual attention of weaker rollouts. GRAD consistently improves performance across two model scales and seven benchmarks, and further complements existing RL methods.

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/2610.21380 — 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 2026-10-16T08:52:02.657Z dalla pagina https://arxiv.org/list/cs.CV/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).