Video-OPSD: Exploiting Privileged Visual Evidence for On-Policy Self-Distillation in Video Large Language Models

• URL: http://arxiv.org/abs/2608.27065v1
• Date: Thu, 27 Aug 2026 12:51:32 GMT
• Title: Video-OPSD: Exploiting Privileged Visual Evidence for On-Policy Self-Distillation in Video Large Language Models
• Abstract summary: On-policy self-distillation (OPSD) has recently emerged as an effective post-training paradigm that improves policy optimization through dense token-level supervision from a privileged self-teacher.<n>We present $textbfVideo-OPSD$, an OPSD framework that exploits privileged visual evidence for both self-teacher construction and knowledge transfer.
• Score: 58.45264308304815
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: On-policy self-distillation (OPSD) has recently emerged as an effective post-training paradigm that improves policy optimization through dense token-level supervision from a privileged self-teacher. Despite its promise, OPSD remains largely underexplored for Video Large Language Models (Video-LLMs). Existing methods typically construct privileged teachers by augmenting their context with additional information while keeping the primary input unchanged for both teacher and student. Video reasoning, however, offers a distinct source of privileged supervision within the primary input itself: long videos contain substantial temporal redundancy, and only a small subset of frames provides the evidence necessary to answer a question. Building on this observation, we present $\textbf{Video-OPSD}$, an OPSD framework that exploits privileged visual evidence for both self-teacher construction and knowledge transfer. First, our Evidence-Grounded Self-Teacher conditions the teacher exclusively on annotated evidence frames while the student continues to reason over the complete video. This focused visual input enables the teacher to provide more informative supervision. Second, our Evidence-Guided Token Optimization adaptively weights token-level distillation according to each reasoning token's reliance on privileged visual evidence, thereby emphasizing perceptually grounded reasoning. Experiments across video understanding and reasoning benchmarks show that $\textbf{Video-OPSD}$ consistently improves upon Standard OPSD across multiple backbones and achieves performance comparable to GRPO while requiring substantially less training time, establishing an effective and efficient post-training approach for Video-LLMs.

Related papers

• Video-MOPD: Multi-Teacher On-Policy Distillation for Video Understanding [32.55965761036089]
  Video understanding demands a convergence of complementary capabilities across perception, temporal understanding, and complex reasoning.<n>We introduce Video-MOPD-8B, an open-weight model dedicated to video understanding tasks.<n>We conduct targeted reinforcement learning (RL) optimization across three core domains: video temporal grounding (VTG), general video comprehension, and video STEM reasoning.
  arXiv  Detail & Related papers  (2026-09-08T18:00:22Z)
• Deep Thought Alignment: Trajectory-Level Latent Distillation for Video Reasoning [42.05846352410443]
  Large Multimodal Models (LMMs) for video reasoning have long been hindered by the high computational cost of processing vast amounts of visual information.<n>We propose Latent-OPD, which augments On-Policy Distillation with trajectory-level latent distillation.<n> Experiments on six video reasoning benchmarks show that Latent-OPD consistently outperforms output-only OPD.
  arXiv  Detail & Related papers  (2026-08-17T09:23:49Z)
• Latent On-Policy Self-Distillation [55.57800223145407]
  On-policy self-distillation (OPSD) offers an effective pathway by using a privileged self-teacher to provide dense supervision on the student's own trajectories.<n>Existing methods still rely heavily on designer-specified privileged artifacts.<n>We introduce Latent On-Policy Self-Distillation (LOPD), which makes the teacher's privileged context itself learnable end-to-end from experience.
  arXiv  Detail & Related papers  (2026-08-13T10:05:51Z)
• VisionCoach: Reinforcing Grounded Video Reasoning via Visual-Perception Prompting [58.508924874097715]
  VisonCoach consists of two components: (1) Visual Prompt Selector, which predicts appropriate prompt types conditioned on the video and question, and (2) Spatio-Reasoner, optimized with RL under visual prompt guidance and object grounding rewards.<n>Our results show that visual prompting during training improves grounded video reasoning, while distract-distillationSense enables the model to this ability without requiring prompts at inference time.
  arXiv  Detail & Related papers  (2026-03-15T23:32:02Z)
• Video-OPD: Efficient Post-Training of Multimodal Large Language Models for Temporal Video Grounding via On-Policy Distillation [29.755136665244805]
  Reinforcement learning has emerged as a principled post-training paradigm for Temporal Video Grounding (TVG)<n>We propose Video-OPD, an efficient post-training framework for TVG inspired by recent advances in on-policy distillation.
  arXiv  Detail & Related papers  (2026-02-03T02:05:48Z)
• Video-LMM Post-Training: A Deep Dive into Video Reasoning with Large Multimodal Models [78.32948112203228]
  Video understanding represents the most challenging frontier in computer vision.<n>Recent emergence of Video-Large Multitemporal Models has demonstrated remarkable capabilities in video understanding tasks.<n>Survey aims to provide researchers and practitioners with a unified framework for advancing Video-LMM capabilities.
  arXiv  Detail & Related papers  (2025-10-06T17:10:44Z)
• Context Consistency Learning via Sentence Removal for Semi-Supervised Video Paragraph Grounding [9.280423086981703]
  We propose a novel Context Consistency Learning (CCL) framework to enhance semi-supervised learning.<n>CCL unifies the paradigms of consistency regularization and pseudo-labeling to enhance semi-supervised learning.
  arXiv  Detail & Related papers  (2025-06-23T10:22:46Z)
• STEP: Enhancing Video-LLMs' Compositional Reasoning by Spatio-Temporal Graph-guided Self-Training [87.58996020705258]
  Video Large Language Models (Video-LLMs) have recently shown strong derivation in basic video understanding tasks.<n>Video-LLMs struggle with compositional reasoning that requires multi-step explicit-temporal inference across object relations, interactions and events.<n>We propose STEP, a novel graph-guided self-training method that enables VideoLLMs to generate reasoning-rich finetuning data from any raw videos to improve itself.
  arXiv  Detail & Related papers  (2024-11-29T11:54:55Z)
• Siamese Learning with Joint Alignment and Regression for Weakly-Supervised Video Paragraph Grounding [70.31050639330603]
  Video paragraph grounding aims at localizing multiple sentences with semantic relations and temporal order from an untrimmed video. Existing VPG approaches are heavily reliant on a considerable number of temporal labels that are laborious and time-consuming to acquire. We introduce and explore Weakly-Supervised Video paragraph Grounding (WSVPG) to eliminate the need of temporal annotations.
  arXiv  Detail & Related papers  (2024-03-18T04:30:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.