Paper Detail
Shravan Venkatraman, Omkar Thawakar, Ritesh Thawkar, Abdelrahman Shaker, Rao Muhammad Anwer
Self-improvement for multimodal large language models (MLLMs) is typically driven by reward-based methods that provide only coarse scalar feedback. Distillation offers a richer alternative through dense token-level supervision, but in the visual domain it usually depends on privileged context constructed using external annotations and tools, or stronger models. We introduce \textbf{CVPD} (Contrastive Counterfactual Visual Process Distillation), which, to the best of our knowledge, is the first fully self-contained framework for dense, on-policy, token-level visual self-distillation for MLLMs. CVPD identifies visual blind spots where zooming into a region changes and sharpens the model's answer distribution, while removing the same region leaves the full-image behavior largely unchanged. Such regions reveal perceptual information that the model can encode but fails to consistently utilize under full-image conditioning. We propose a three-gate Counterfactual Criterion that identifies these regions directly from the model's own responses and converts them into dense contrastive supervision for self-distillation. On Qwen3-VL-8B-Instruct, CVPD outperforms six self-evolving baselines across twelve benchmarks, including methods that rely on external GPT-4o supervision, without a single regression. It achieves gains of $+3.60$ on OCRBench, $+3.38$ on MMStar Fine-Grained Perception, and $+3.08$ on MMStar Logical Reasoning, while maintaining or improving performance on broader multimodal benchmarks.
No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.
No ranking explanation is available yet.
No tags.
@article{venkatraman2026perception,
title = {Perception Before Supervision: Self-Contained Visual Distillation from Counterfactual Blind Spots},
author = {Shravan Venkatraman and Omkar Thawakar and Ritesh Thawkar and Abdelrahman Shaker and Rao Muhammad Anwer},
year = {2026},
abstract = {Self-improvement for multimodal large language models (MLLMs) is typically driven by reward-based methods that provide only coarse scalar feedback. Distillation offers a richer alternative through dense token-level supervision, but in the visual domain it usually depends on privileged context constructed using external annotations and tools, or stronger models. We introduce \textbackslash{}textbf\{CVPD\} (Contrastive Counterfactual Visual Process Distillation), which, to the best of our knowledge, is the first f},
url = {https://arxiv.org/abs/2608.09931},
keywords = {cs.CV},
eprint = {2608.09931},
archiveprefix = {arXiv},
}
{}