Metraipolis › archive › dc_f03bcc1e13eb

CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction — arXiv:2609.02401

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.02401
License (as archived)
CC0-1.0
Archived
2026-10-17T05:50:03.079Z
SHA-256 of the archived text
40603aea49e583fd0290467f9f4a98131bd0185b23601929688692ffa5644089
Stable URL of this page
https://metraipolis.com/documento/dc_f03bcc1e13eb

How to cite

CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction — arXiv:2609.02401 — arXiv (en), https://arxiv.org/abs/2609.02401 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f03bcc1e13eb (Metraipolis archive dc_f03bcc1e13eb, sha256 40603aea49e583fd0290467f9f4a98131bd0185b23601929688692ffa5644089)
BibTeX
@misc{dcf03bcc1e13eb,
  title = {CA-OPD: Confidence-Aware On-Policy Distillation for Structured Visual Prediction — arXiv:2609.02401},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.02401},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f03bcc1e13eb; sha256 40603aea49e583fd0290467f9f4a98131bd0185b23601929688692ffa5644089},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.02401. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Autoregressive vision language models unify heterogeneous perception tasks but are highly susceptible to compounding errors. On-policy distillation (OPD) bridges the training-inference mismatch by training students on their own rollouts. However, unreliable student predictions, especially early in training, can derail the trajectory and degrade the quality of teacher supervision. While recent interleaved distillation methods allow the teacher to verify and replace student tokens, they primarily rely on rigid ranking metrics rather than exact teacher confidence, and they overlook how intervention decisions can inform token-level supervision. To address this, we introduce Confidence-Aware On-Policy Distillation (CA-OPD), a framework that couples reliable rollout construction with adaptive supervision. CA-OPD utilizes teacher confidence to selectively correct unreliable student transitions, gradually transferring rollout control to the student via a strict-to-relaxed schedule. Crucially, CA-OPD aligns knowledge transfer with these intervention decisions: corrected positions receive direct cross-entropy supervision from the teacher's prediction, while retained positions benefit from the teacher's full predictive distribution. Evaluated in a multi-teacher setting for GUI grounding and optical character recognition, CA-OPD substantially improves the Qwen3.5-0.8B baseline across all six target benchmarks, including gains of $9.50$ points on ScreenSpot-Pro and $6.72$ points on OCRBench-v2 English. Controlled studies further show that the gains depend on intervention placement, progressive rollout control, and intervention-aligned supervision, rather than intervention frequency alone.

Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-03 in cs.CV).
Categorie arXiv: Computer Vision and Pattern Recognition (cs.CV).
Pagina del lavoro: https://arxiv.org/abs/2609.02401 — 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 2026-10-17T05:25:58.112Z dalla pagina https://arxiv.org/catchup/cs.CV/2026-09-03?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).