# Keyframe-oriented Vision Token Pruning: Enhancing Efficiency of Large Vision Language Models on Long-Form Video Processing

arXiv: https://arxiv.org/abs/2503.10742 · HTML: https://mlsanity.com/paper/2503.10742/

## Key facts

- Cited by 18 papers in the map (data through 14 Oct 2026).
- Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 32 of its 40 references that are in the map.
- First posted on arXiv 13 Mar 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)
- Efficient LLM inference and serving / [Visual token pruning in multimodal models](https://mlsanity.com/inference/visual-token-pruning-in-multimodal-models/)

## Builds on (most cited first, 30 of 32)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Sigmoid Loss for Language Image Pre-Training](https://mlsanity.com/paper/2303.15343/) (2023)
- [InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuning](https://mlsanity.com/paper/2305.06500/) (2023)
- [Video-LLaMA: An Instruction-tuned Audio-Visual Language Model for Video Understanding](https://mlsanity.com/paper/2306.02858/) (2023)
- [Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis](https://mlsanity.com/paper/2405.21075/) (2024)
- [Video-LLaVA: Learning United Visual Representation by Alignment Before Projection](https://mlsanity.com/paper/2311.10122/) (2023)
- [MVBench: A Comprehensive Multi-modal Video Understanding Benchmark](https://mlsanity.com/paper/2311.17005/) (2023)
- [Video-ChatGPT: Towards Detailed Video Understanding via Large Vision and Language Models](https://mlsanity.com/paper/2306.05424/) (2023)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [ShareGPT4V: Improving Large Multi-Modal Models with Better Captions](https://mlsanity.com/paper/2311.12793/) (2023)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [VideoChat: Chat-Centric Video Understanding](https://mlsanity.com/paper/2305.06355/) (2023)
- [NExT-QA:Next Phase of Question-Answering to Explaining Temporal Actions](https://mlsanity.com/paper/2105.08276/) (2021)
- [EgoSchema: A Diagnostic Benchmark for Very Long-form Video Language Understanding](https://mlsanity.com/paper/2308.09126/) (2023)
- [An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models](https://mlsanity.com/paper/2403.06764/) (2024)
- [ActivityNet-QA: A Dataset for Understanding Complex Web Videos via Question Answering](https://mlsanity.com/paper/1906.02467/) (2019)
- [TimeChat: A Time-sensitive Multimodal Large Language Model for Long Video Understanding](https://mlsanity.com/paper/2312.02051/) (2023)
- [LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models](https://mlsanity.com/paper/2403.15388/) (2024)
- [Perception Test: A Diagnostic Benchmark for Multimodal Video Models](https://mlsanity.com/paper/2305.13786/) (2023)
- [VideoTree: Adaptive Tree-based Video Representation for LLM Reasoning on Long Videos](https://mlsanity.com/paper/2405.19209/) (2024)
- [SPViT: Enabling Faster Vision Transformers via Soft Token Pruning](https://mlsanity.com/paper/2112.13890/) (2021)
- [AuroraCap: Efficient, Performant Video Detailed Captioning and a New Benchmark](https://mlsanity.com/paper/2410.03051/) (2024)
- [Joint Token Pruning and Squeezing Towards More Aggressive Compression of Vision Transformers](https://mlsanity.com/paper/2304.10716/) (2023)
- [Too Many Frames, Not All Useful: Efficient Strategies for Long-Form Video QA](https://mlsanity.com/paper/2406.09396/) (2024)
- [VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection](https://mlsanity.com/paper/2411.14794/) (2024)
- [Koala: Key frame-conditioned long video-LLM](https://mlsanity.com/paper/2404.04346/) (2024)
- [KeyVideoLLM: Towards Large-scale Video Keyframe Selection](https://mlsanity.com/paper/2407.03104/) (2024)
- [vid-TLDR: Training Free Token merging for Light-weight Video Transformer](https://mlsanity.com/paper/2403.13347/) (2024)
- [Zero-Shot Video Moment Retrieval from Frozen Vision-Language Models](https://mlsanity.com/paper/2309.00661/) (2023)

## Built on by (most cited first, 18 of 18)

- [Video-MMLU: A Massive Multi-Discipline Lecture Understanding Benchmark](https://mlsanity.com/paper/2504.14693/) (2025)
- [Angles Don't Lie: Unlocking Training-Efficient RL Through the Model's Own Signals](https://mlsanity.com/paper/2506.02281/) (2025)
- [CoreMatching: A Co-adaptive Sparse Inference Framework with Token and Neuron Pruning for Comprehensive Acceleration of Vision-Language Models](https://mlsanity.com/paper/2505.19235/) (2025)
- [Language-Guided Temporal Token Pruning for Efficient VideoLLM Processing](https://mlsanity.com/paper/2508.17686/) (2025)
- [SeViCES: Unifying Semantic-Visual Evidence Consensus for Long Video Understanding](https://mlsanity.com/paper/2510.20622/) (2025)
- [Towards Fast and Effective Long Video Understanding of Multimodal Large Language Models via Adaptive Quasi-Gaussian Sampling](https://mlsanity.com/paper/2606.24187/) (2026)
- [Efficient Video Sampling: Pruning Temporally Redundant Tokens for Faster VLM Inference](https://mlsanity.com/paper/2510.14624/) (2025)
- [Geometry-Guided 3D Visual Token Pruning for Video-Language Models](https://mlsanity.com/paper/2604.18260/) (2026)
- [Dynamic Hub-and-Spoke Memory for Streaming Video Understanding](https://mlsanity.com/paper/2608.30294/) (2026)
- [Focus: A Streaming Concentration Architecture for Efficient Vision-Language Models](https://mlsanity.com/paper/2512.14661/) (2025)
- [MemoryCard: Topic-Aware Multi-Modal Clue Compression for Long-Video Question Answering](https://mlsanity.com/paper/2606.05917/) (2026)
- [One Ranking, Any Budget: Matryoshka Evidence-to-Context Frame Selection for Long-Video Understanding](https://mlsanity.com/paper/2608.05707/) (2026)
- [Baseline-Relative Counterfactual Refinement for Bit-Aware Visual Token Communication](https://mlsanity.com/paper/2608.16192/) (2026)
- [ConsensusDrop: Fusing Visual and Cross-Modal Saliency for Efficient Vision Language Models](https://mlsanity.com/paper/2602.00946/) (2026)
- [Development of Vision-Language Model-based GNSS Spoofing Detection for Autonomous Vehicle Navigation](https://mlsanity.com/paper/2607.23962/) (2026)
- [GSTEP: Global Spatio-Temporal Density-Driven Visual Token Pruning for Efficient Video Large Language Models](https://mlsanity.com/paper/2608.03083/) (2026)
- [Coarse Indexing, Fine Evidence: Decoupling Temporal Granularity in Long-Video RAG](https://mlsanity.com/paper/2608.23011/) (2026)
- [KeyRec: Bounded Visual Memory for Streaming and Long-Video Understanding](https://mlsanity.com/paper/2609.32182/) (2026)

## Abstract (arXiv)

Vision language models (VLMs) demonstrate strong capabilities in jointly processing visual and textual data. However, they often incur substantial computational overhead due to redundant visual information, particularly in long-form video scenarios. Existing approaches predominantly focus on either vision token pruning, which may overlook spatio-temporal dependencies, or keyframe selection, which identifies informative frames but discards others, thus disrupting contextual continuity. In this work, we propose KVTP (Keyframe-oriented Vision Token Pruning), a novel framework that overcomes the drawbacks of token pruning and keyframe selection. By adaptively assigning pruning rates based on frame relevance to the query, KVTP effectively retains essential contextual information while significantly reducing redundant computation. To thoroughly evaluate the long-form video understanding capacities of VLMs, we curated and reorganized subsets from VideoMME, EgoSchema, and NextQA into a unified benchmark named SparseKV-QA that highlights real-world scenarios with sparse but crucial events. Our experiments with VLMs of various scales show that KVTP can reduce token usage by 80% without compromising spatiotemporal and contextual consistency, significantly cutting computation while maintaining the performance. These results demonstrate our approach's effectiveness in efficient long-video processing, facilitating more scalable VLM deployment.