# Evidence-Aligned Multimodal On-Policy Self-Distillation for Fine-Grained Visual Understanding

arXiv: https://arxiv.org/abs/2609.34672 · HTML: https://mlsanity.com/paper/2609.34672/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Multimodal distillation, one of 12 directions in Knowledge distillation.
- Builds on 34 of its 38 references that are in the map.
- First posted on arXiv 28 Sep 2026.

## Placed in

- Knowledge distillation / [Multimodal distillation](https://mlsanity.com/distill/multimodal-distillation/)

## Builds on (most cited first, 30 of 34)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes](https://mlsanity.com/paper/2306.13649/) (2023)
- [Are We on the Right Way for Evaluating Large Vision-Language Models?](https://mlsanity.com/paper/2403.20330/) (2024)
- [Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs](https://mlsanity.com/paper/2406.16860/) (2024)
- [Sequence-Level Knowledge Distillation](https://mlsanity.com/paper/1606.07947/) (2016)
- [Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs](https://mlsanity.com/paper/2401.06209/) (2024)
- [Mitigating Object Hallucinations in Large Vision-Language Models through Visual Contrastive Decoding](https://mlsanity.com/paper/2311.16922/) (2023)
- [V*: Guided Visual Search as a Core Mechanism in Multimodal LLMs](https://mlsanity.com/paper/2312.14135/) (2023)
- [GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning](https://mlsanity.com/paper/2507.01006/) (2025)
- [Self-Distilled Reasoner: On-Policy Self-Distillation for Large Language Models](https://mlsanity.com/paper/2601.18734/) (2026)
- [DeepEyes: Incentivizing "Thinking with Images" via Reinforcement Learning](https://mlsanity.com/paper/2505.14362/) (2025)
- [MME-RealWorld: Could Your Multimodal LLM Challenge High-Resolution Real-World Scenarios that are Difficult for Humans?](https://mlsanity.com/paper/2408.13257/) (2024)
- [MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe](https://mlsanity.com/paper/2509.18154/) (2025)
- [MLLMs Know Where to Look: Training-free Perception of Small Visual Details with Multimodal LLMs](https://mlsanity.com/paper/2502.17422/) (2025)
- [Thyme: Think Beyond Images](https://mlsanity.com/paper/2508.11630/) (2025)
- [Divide, Conquer and Combine: A Training-Free Framework for High-Resolution Image Perception in Multimodal Large Language Models](https://mlsanity.com/paper/2408.15556/) (2024)
- [DeepEyesV2: Toward Agentic Multimodal Model](https://mlsanity.com/paper/2511.05271/) (2025)
- [MiMo-VL Technical Report](https://mlsanity.com/paper/2506.03569/) (2025)
- [Vision-OPD: Learning to See Fine Details for Multimodal LLMs via On-Policy Self-Distillation](https://mlsanity.com/paper/2605.18740/) (2026)
- [Zooming without Zooming: Region-to-Image Distillation for Fine-Grained Multimodal Perception](https://mlsanity.com/paper/2602.11858/) (2026)
- [Selective Knowledge Distillation for Neural Machine Translation](https://mlsanity.com/paper/2105.12967/) (2021)
- [SenseNova-MARS: Empowering Multimodal Agentic Reasoning and Search via Reinforcement Learning](https://mlsanity.com/paper/2512.24330/) (2025)
- [Visual-Advantage On-Policy Distillation for Vision-Language Models](https://mlsanity.com/paper/2605.21924/) (2026)
- [Grasp Any Region: Towards Precise, Contextual Pixel Understanding for Multimodal LLMs](https://mlsanity.com/paper/2510.18876/) (2025)
- [VLM-FO1: Bridging the Gap Between High-Level Reasoning and Fine-Grained Perception in VLMs](https://mlsanity.com/paper/2509.25916/) (2025)
- [V-Zero: Answer-Label-Free On-Policy Distillation with Contrastive Evidence Gating for Fine-Grained Visual Reasoning](https://mlsanity.com/paper/2606.25319/) (2026)
- [Rethinking Selective Knowledge Distillation](https://mlsanity.com/paper/2602.01395/) (2026)
- [OPD-V: Visual On-Policy Self-Distillation with Modality Balance](https://mlsanity.com/paper/2608.05131/) (2026)

## Abstract (arXiv)

Fine-grained visual understanding requires models to recognize small details within complex images. Multimodal on-policy self-distillation (OPSD) addresses this challenge by using a teacher conditioned on evidence-centered crops to supervise a student conditioned on original images along student-generated trajectories. Ideally, teacher corrections, the distributional changes from the student toward the privileged teacher, should be driven by task-relevant visual evidence. However, the designs that make the teacher effective also introduce other interference. Using a lagged or frozen teacher improves training stability but introduces a model-state gap from the evolving student, while cropping enhances task-relevant evidence but also loses the visual context. These two sources of interference make the teacher corrections not purely rely on the visual evidence. We introduce Evidence-Aligned multimodal on-policy self-Distillation (EAD), which retains the crop-conditioned teacher as the target but constructs a separate evidence reference for weighting the corrections. To exclude the effect of lagged model-state from this reference, EAD measures prediction changes using the current student. To avoid crop-induced context changes, EAD masks the evidence region in the original image while preserving the other visual context. The change from the student's masked-image prediction to its original-image prediction provides a controlled reference for the direction in which the visual evidence shifts the student's prediction. EAD weights each teacher correction by its cosine alignment with the reference, i.e., retaining aligned corrections and downweighting the rest. Retaining only 6% of the supervision mass of dense OPSD, EAD consistently outperforms previous state-of-the-art methods.