# Beyond Training: Dynamic Token Merging for Zero-Shot Video Understanding

arXiv: https://arxiv.org/abs/2411.14401 · HTML: https://mlsanity.com/paper/2411.14401/

## Key facts

- Cited by 13 papers in the map (data through 14 Oct 2026).
- Placed in Video understanding with MLLMs, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 37 of its 46 references that are in the map.
- First posted on arXiv 21 Nov 2024.

## Placed in

- Multimodal LLMs / vision-language models / [Video understanding with MLLMs](https://mlsanity.com/mllm/video-understanding-with-mllms/)
- 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 37)

- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [Is Space-Time Attention All You Need for Video Understanding?](https://mlsanity.com/paper/2102.05095/) (2021)
- [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)
- [Video Swin Transformer](https://mlsanity.com/paper/2106.13230/) (2021)
- [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)
- [TSM: Temporal Shift Module for Efficient Video Understanding](https://mlsanity.com/paper/1811.08383/) (2018)
- [VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs](https://mlsanity.com/paper/2406.07476/) (2024)
- [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)
- [ActivityNet-QA: A Dataset for Understanding Complex Web Videos via Question Answering](https://mlsanity.com/paper/1906.02467/) (2019)
- [LLaMA-VID: An Image is Worth 2 Tokens in Large Language Models](https://mlsanity.com/paper/2311.17043/) (2023)
- [MovieChat: From Dense Token to Sparse Memory for Long Video Understanding](https://mlsanity.com/paper/2307.16449/) (2023)
- [PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning](https://mlsanity.com/paper/2404.16994/) (2024)
- [InternVideo2: Scaling Foundation Models for Multimodal Video Understanding](https://mlsanity.com/paper/2403.15377/) (2024)
- [MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training](https://mlsanity.com/paper/2403.09611/) (2024)
- [VLM2Vec: Training Vision-Language Models for Massive Multimodal Embedding Tasks](https://mlsanity.com/paper/2410.05160/) (2024)
- [STAR: A Benchmark for Situated Reasoning in Real-World Videos](https://mlsanity.com/paper/2405.09711/) (2024)
- [Video Understanding with Large Language Models: A Survey](https://mlsanity.com/paper/2312.17432/) (2023)
- [Self-Chained Image-Language Model for Video Localization and Question Answering](https://mlsanity.com/paper/2305.06988/) (2023)
- [UniFormer: Unified Transformer for Efficient Spatiotemporal Representation Learning](https://mlsanity.com/paper/2201.04676/) (2022)
- [HALC: Object Hallucination Reduction via Adaptive Focal-Contrast Decoding](https://mlsanity.com/paper/2403.00425/) (2024)
- [SlowFast-LLaVA: A Strong Training-Free Baseline for Video Large Language Models](https://mlsanity.com/paper/2407.15841/) (2024)
- [TGIF: A New Dataset and Benchmark on Animated GIF Description](https://mlsanity.com/paper/1604.02748/) (2016)
- [Evaluating Open-Domain Question Answering in the Era of Large Language Models](https://mlsanity.com/paper/2305.06984/) (2023)
- [Efficient Parameter-free Clustering Using First Neighbor Relations](https://mlsanity.com/paper/1902.11266/) (2019)
- [An Image Grid Can Be Worth a Video: Zero-shot Video Question Answering Using a VLM](https://mlsanity.com/paper/2403.18406/) (2024)
- [Ferret-v2: An Improved Baseline for Referring and Grounding with Large Language Models](https://mlsanity.com/paper/2404.07973/) (2024)

## Built on by (most cited first, 13 of 13)

- [StreamMem: Query-Agnostic KV Cache Memory for Streaming Video Understanding](https://mlsanity.com/paper/2508.15717/) (2025)
- [MemFlow: Flowing Adaptive Memory for Consistent and Efficient Long Video Narratives](https://mlsanity.com/paper/2512.14699/) (2025)
- [CAFe: Unifying Representation and Generation with Contrastive-Autoregressive Finetuning](https://mlsanity.com/paper/2503.19900/) (2025)
- [Universal Video Temporal Grounding with Generative Multi-modal Large Language Models](https://mlsanity.com/paper/2506.18883/) (2025)
- [Efficient Multi-modal Large Language Models via Progressive Consistency Distillation](https://mlsanity.com/paper/2510.00515/) (2025)
- [KTV: Keyframes and Key Tokens Selection for Efficient Training-Free Video LLMs](https://mlsanity.com/paper/2602.03615/) (2026)
- [AdaCluster: Adaptive Query-Key Clustering for Sparse Attention in Video Generation](https://mlsanity.com/paper/2604.18348/) (2026)
- [video-SALMONN S: Memory-Enhanced Streaming Audio-Visual LLM](https://mlsanity.com/paper/2510.11129/) (2025)
- [Stay Unique, Stay Efficient: Preserving Model Personality in Multi-Task Merging](https://mlsanity.com/paper/2512.01461/) (2025)
- [A Multi-Agent Perception-Action Alliance for Efficient Long Video Reasoning](https://mlsanity.com/paper/2603.14052/) (2026)
- [CircuitProbe: Tracing Visual Temporal Evidence Flow in Video Language Models](https://mlsanity.com/paper/2507.19420/) (2025)
- [GRIP-VLM: Group-Relative Importance Pruning for Efficient Vision-Language Models](https://mlsanity.com/paper/2605.13375/) (2026)
- [video-SALMONN-R3: Learning to ReWatch, ReAsk, and ReAnswer for Efficient Video Understanding](https://mlsanity.com/paper/2606.24477/) (2026)

## Abstract (arXiv)

Recent advancements in multimodal large language models (MLLMs) have opened new avenues for video understanding. However, achieving high fidelity in zero-shot video tasks remains challenging. Traditional video processing methods rely heavily on fine-tuning to capture nuanced spatial-temporal details, which incurs significant data and computation costs. In contrast, training-free approaches, though efficient, often lack robustness in preserving context-rich features across complex video content. To this end, we propose DYTO, a novel dynamic token merging framework for zero-shot video understanding that adaptively optimizes token efficiency while preserving crucial scene details. DYTO integrates a hierarchical frame selection and a bipartite token merging strategy to dynamically cluster key frames and selectively compress token sequences, striking a balance between computational efficiency with semantic richness. Extensive experiments across multiple benchmarks demonstrate the effectiveness of DYTO, achieving superior performance compared to both fine-tuned and training-free methods and setting a new state-of-the-art for zero-shot video understanding.