# Dynamic Token Reduction during Generation for Vision Language Models

arXiv: https://arxiv.org/abs/2501.14204 · HTML: https://mlsanity.com/paper/2501.14204/

## Key facts

- Cited by 10 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 6 of its 30 references that are in the map.
- First posted on arXiv 24 Jan 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, 6 of 6)

- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [Video-LLaVA: Learning United Visual Representation by Alignment Before Projection](https://mlsanity.com/paper/2311.10122/) (2023)
- [SparseVLM: Visual Token Sparsification for Efficient Vision-Language Model Inference](https://mlsanity.com/paper/2410.04417/) (2024)
- [VisionZip: Longer is Better but Not Necessary in Vision Language Models](https://mlsanity.com/paper/2412.04467/) (2024)
- [LLaVA-Plus: Learning to Use Tools for Creating Multimodal Agents](https://mlsanity.com/paper/2311.05437/) (2023)
- [TokenPacker: Efficient Visual Projector for Multimodal LLM](https://mlsanity.com/paper/2407.02392/) (2024)

## Built on by (most cited first, 10 of 10)

- [A Glimpse to Compress: Dynamic Visual Token Pruning for Large Vision-Language Models](https://mlsanity.com/paper/2508.01548/) (2025)
- [DyMU: Dynamic Merging and Virtual Unmerging for Efficient VLMs](https://mlsanity.com/paper/2504.17040/) (2025)
- [Toward Native Multimodal Modeling: A Roadmap](https://mlsanity.com/paper/2605.25343/) (2026)
- [Beyond Surrogate Gradients: Fully Differentiable Token Pruning for Vision-Language Models](https://mlsanity.com/paper/2605.28051/) (2026)
- [Accelerating Multimodal Large Language Models with Prior-Corrected Token Reduction](https://mlsanity.com/paper/2606.24156/) (2026)
- [GreedyPrune: Retenting Critical Visual Token Set for Large Vision Language Models](https://mlsanity.com/paper/2506.13166/) (2025)
- [Spectral Evolution-Guided Token Pruning in Multimodal Large Language Models](https://mlsanity.com/paper/2606.24165/) (2026)
- [DiffPrune: differentiable information throttling for token pruning in vision-language models](https://mlsanity.com/paper/2608.01985/) (2026)
- [Improving MLLM Training Efficiency via Stage-Aware Sparsity](https://mlsanity.com/paper/2509.18150/) (2025)
- [Attending to Multimodal Generation One Token at a Time](https://mlsanity.com/paper/2607.03738/) (2026)

## Abstract (arXiv)

Vision-Language Models (VLMs) have achieved notable success in multimodal tasks but face practical limitations due to the quadratic complexity of decoder attention mechanisms and autoregressive generation. Existing methods like FASTV and VTW have achieved notable results in reducing redundant visual tokens, but these approaches focus on pruning tokens in a single forward pass without systematically analyzing the redundancy of visual tokens throughout the entire generation process. In this paper, we introduce a dynamic pruning strategy tailored for VLMs, namedDynamic Rate (DyRate), which progressively adjusts the compression rate during generation. Our analysis of the distribution of attention reveals that the importance of visual tokens decreases throughout the generation process, inspiring us to adopt a more aggressive compression rate. By integrating a lightweight predictor based on attention distribution, our approach enables flexible adjustment of pruning rates based on the attention distribution. Our experimental results demonstrate that our method not only reduces computational demands but also maintains the quality of responses.