TSPO: Temporal Sampling Policy Optimization for Long-form Video Language Understanding

• URL: http://arxiv.org/abs/2508.04369v2
• Date: Thu, 07 Aug 2025 15:24:17 GMT
• Title: TSPO: Temporal Sampling Policy Optimization for Long-form Video Language Understanding
• Authors: Canhui Tang, Zifan Han, Hongbo Sun, Sanping Zhou, Xuchong Zhang, Xin Wei, Ye Yuan, Jinglin Xu, Hao Sun,
• Abstract summary: Multi-language Large Language Models (MLLMs) have demonstrated significant progress in vision-based tasks, yet they still face challenges when processing long-duration video inputs.<n>We propose Temporal Policy Sampling Optimization (TSPO), advancing MLLMs' long-form video-language understanding via reinforcement learning.<n>Our TSPO achieves state-of-the-art performance across multiple long video understanding benchmarks, and shows transferable ability across different cutting-edge Video-MLLMs.
• Score: 26.463523465270097
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multimodal Large Language Models (MLLMs) have demonstrated significant progress in vision-language tasks, yet they still face challenges when processing long-duration video inputs. The limitation arises from MLLMs' context limit and training costs, necessitating sparse frame sampling before feeding videos into MLLMs. Existing video MLLMs adopt training-free uniform sampling or keyframe search, which may miss critical events or be constrained by the pre-trained models' event understanding capabilities. Meanwhile, building a training-based method remains challenging due to the unsupervised and non-differentiable nature of sparse frame sampling. To address these problems, we propose Temporal Sampling Policy Optimization (TSPO), advancing MLLMs' long-form video-language understanding via reinforcement learning. Specifically, we first propose a trainable event-aware temporal agent, which captures event-query correlation for performing probabilistic keyframe selection. Then, we propose the TSPO reinforcement learning paradigm, which models keyframe selection and language generation as a joint decision-making process, enabling end-to-end group relative optimization with efficient rule-based rewards. Furthermore, for the TSPO's training, we propose a long video training data construction pipeline with comprehensive temporal data and video Needle-in-a-Haystack data. Finally, we incorporate rule-based answering accuracy and temporal locating reward mechanisms to optimize the temporal sampling policy. Comprehensive experiments show that our TSPO achieves state-of-the-art performance across multiple long video understanding benchmarks, and shows transferable ability across different cutting-edge Video-MLLMs. Our code is available at https://github.com/Hui-design/TSPO

Related papers

• How Important are Videos for Training Video LLMs? [55.965474658745315]
  We present findings indicating Video LLMs are more capable of temporal reasoning after image-only training than one would assume.<n>We introduce a simple finetuning scheme involving sequences of annotated images and questions targeting temporal capabilities.<n>This suggests suboptimal utilization of rich temporal features found in real video by current models.
  arXiv  Detail & Related papers  (2025-06-07T21:32:19Z)
• Reinforcement Learning Tuning for VideoLLMs: Reward Design and Data Efficiency [56.475612147721264]
  We propose a dual-reward formulation that supervises both semantic and temporal reasoning through discrete and continuous reward signals.<n>We evaluate our approach across eight representative video understanding tasks, including VideoQA, Temporal Video Grounding, and Grounded VideoQA.<n>Results underscore the importance of reward design and data selection in advancing reasoning-centric video understanding with MLLMs.
  arXiv  Detail & Related papers  (2025-06-02T17:28:26Z)
• ViaRL: Adaptive Temporal Grounding via Visual Iterated Amplification Reinforcement Learning [68.76048244253582]
  We introduce ViaRL, the first framework to leverage rule-based reinforcement learning (RL) for optimizing frame selection in video understanding.<n>ViaRL utilizes the answer accuracy of a downstream model as a reward signal to train a frame selector through trial-and-error.<n>ViaRL consistently delivers superior temporal grounding performance and robust generalization across diverse video understanding tasks.
  arXiv  Detail & Related papers  (2025-05-21T12:29:40Z)
• TEMPLE:Temporal Preference Learning of Video LLMs via Difficulty Scheduling and Pre-SFT Alignment [48.94844127553743]
  TEMPLE is a systematic framework that enhances temporal reasoning capabilities of Video Large Language Models.<n>Our approach consistently improves Video LLM performance across multiple benchmarks with a relatively small set of self-generated DPO data.<n>Our findings highlight our TEMPLE as a scalable and efficient complement to SFT-based methods, paving the way for developing reliable Video LLMs.
  arXiv  Detail & Related papers  (2025-03-21T08:00:29Z)
• Free Video-LLM: Prompt-guided Visual Perception for Efficient Training-free Video LLMs [56.040198387038025]
  We present a novel prompt-guided visual perception framework (abbreviated as Free Video-LLM) for efficient inference of training-free video LLMs. Our method effectively reduces the number of visual tokens while maintaining high performance across multiple video question-answering benchmarks.
  arXiv  Detail & Related papers  (2024-10-14T12:35:12Z)
• Training-free Video Temporal Grounding using Large-scale Pre-trained Models [41.71055776623368]
  Video temporal grounding aims to identify video segments within untrimmed videos that are most relevant to a given natural language query. Existing video temporal localization models rely on specific datasets for training and have high data collection costs. We propose a Training-Free Video Temporal Grounding approach that leverages the ability of pre-trained large models.
  arXiv  Detail & Related papers  (2024-08-29T02:25:12Z)
• LifelongMemory: Leveraging LLMs for Answering Queries in Long-form Egocentric Videos [15.127197238628396]
  LifelongMemory is a new framework for accessing long-form egocentric videographic memory through natural language question answering and retrieval. Our approach achieves state-of-the-art performance on the benchmark for question answering and is highly competitive on the natural language query (NLQ) challenge of Ego4D.
  arXiv  Detail & Related papers  (2023-12-07T19:19:25Z)
• MIST: Multi-modal Iterative Spatial-Temporal Transformer for Long-form Video Question Answering [73.61182342844639]
  We introduce a new model named Multi-modal Iterative Spatial-temporal Transformer (MIST) to better adapt pre-trained models for long-form VideoQA. MIST decomposes traditional dense spatial-temporal self-attention into cascaded segment and region selection modules. Visual concepts at different granularities are then processed efficiently through an attention module.
  arXiv  Detail & Related papers  (2022-12-19T15:05:40Z)

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.