Skip to content
AI News HubLIVE
Original source2 min read

VisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation

Summary

VisAlloc introduces a dual-level framework that combines sample-level visual filtering with token-level counterfactual reweighting to improve visual reasoning in multimodal large language models through on-policy self-distillation.

SourcearXiv Computer VisionAuthor: Haoran Xu, Tian Chen, Yuhe Wu, Yanbo Zhao, Junchi Liao, Hongyu Wang, Jiaze Li, Wentao Wang, Yanfeng Li, Rui Qian
VisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 18 Nov 2026]

Title:VisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation

View a PDF of the paper titled VisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation, by Haoran Xu and 9 other authors

View PDF HTML (experimental)

Abstract:On-policy self-distillation (OPSD) offers a promising approach to improving visual reasoning in multimodal large language models (MLLMs). However, not all samples benefit equally from privileged visual guidance, and uniform token supervision can dilute visually grounded signals. We propose VisAlloc, a dual-level framework combining sample-level visual filtering with token-level counterfactual reweighting. A fixed paired-teacher filter retains samples answered correctly from an evidence-centered positive crop but incorrectly from a spatially displaced negative crop. At the token level, both EMA teachers score the same student-generated response, and their absolute log-probability differences measure view sensitivity. Additional supervision targets tokens admitted by sensitivity or teacher-distribution disagreement, with shared frozen-prior, template-bigram, and student--teacher divergence gates. Inverse-frequency redistribution preserves the total additional weight within each micro-batch, while the positive teacher remains the sole distillation target. Inference requires only the full-image student. Across eight benchmarks, our 4B and 9B models achieve macro-averages of 81.44 and 84.92, outperforming Vision-OPD reproductions by 0.97 and 0.52 points, respectively. Code on GitHub: this https URL

Subjects:

Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)

Cite as: arXiv:2611.25069 [cs.CV]

(or arXiv:2611.25069v1 [cs.CV] for this version)

https://doi.org/10.48550/arXiv.2611.25069

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Tian Chen [view email] [v1] Wed, 18 Nov 2026 19:12:00 UTC (3,812 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled VisAlloc: Allocating Visual Supervision for On-Policy Self-Distillation, by Haoran Xu and 9 other authors

View PDF

HTML (experimental)

TeX Source

view license

Additional Features

Audio Summary

Current browse context:

cs.CV

new | recent | 2026-11

Change to browse by:

cs cs.AI

References & Citations

SciX

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersAdvanced

Key points

  • Sample-level filtering retains samples answered correctly from a positive crop but incorrectly from a negative crop.
  • Token-level reweighting uses log-probability differences between EMA teachers to measure view sensitivity.
  • 4B and 9B models achieve macro-averages of 81.44 and 84.92, outperforming Vision-OPD reproductions.

Highlights and analysis are generated automatically and may contain errors. Check the original source.