# TokenCarve: Information-Preserving Visual Token Compression in Multimodal Large Language Models

arXiv: https://arxiv.org/abs/2503.10501 · HTML: https://mlsanity.com/paper/2503.10501/

## Key facts

- Cited by 19 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 37 of its 44 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 37)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answering](https://mlsanity.com/paper/1612.00837/) (2016)
- [MMBench: Is Your Multi-modal Model an All-around Player?](https://mlsanity.com/paper/2307.06281/) (2023)
- [Learn to Explain: Multimodal Reasoning via Thought Chains for Science Question Answering](https://mlsanity.com/paper/2209.09513/) (2022)
- [Evaluating Object Hallucination in Large Vision-Language Models](https://mlsanity.com/paper/2305.10355/) (2023)
- [Towards VQA Models That Can Read](https://mlsanity.com/paper/1904.08920/) (2019)
- [MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models](https://mlsanity.com/paper/2306.13394/) (2023)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities](https://mlsanity.com/paper/2308.02490/) (2023)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [A-OKVQA: A Benchmark for Visual Question Answering using World Knowledge](https://mlsanity.com/paper/2206.01718/) (2022)
- [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)
- [LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models](https://mlsanity.com/paper/2403.15388/) (2024)
- [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)
- [PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction](https://mlsanity.com/paper/2410.17247/) (2024)
- [LLaVA-Mini: Efficient Image and Video Large Multimodal Models with One Vision Token](https://mlsanity.com/paper/2501.03895/) (2025)
- [DeepStack: Deeply Stacking Visual Tokens is Surprisingly Simple and Effective for LMMs](https://mlsanity.com/paper/2406.04334/) (2024)
- [Multi-Stage Vision Token Dropping: Towards Efficient Multimodal Large Language Model](https://mlsanity.com/paper/2411.10803/) (2024)
- [Dynamic-LLaVA: Efficient Multimodal Large Language Models via Dynamic Vision-language Context Sparsification](https://mlsanity.com/paper/2412.00876/) (2024)
- [MADTP: Multimodal Alignment-Guided Dynamic Token Pruning for Accelerating Vision-Language Transformer](https://mlsanity.com/paper/2403.02991/) (2024)
- [Efficient Large Multi-modal Models via Visual Context Compression](https://mlsanity.com/paper/2406.20092/) (2024)

## Built on by (most cited first, 19 of 19)

- [VScan: Rethinking Visual Token Reduction for Efficient Large Vision-Language Models](https://mlsanity.com/paper/2505.22654/) (2025)
- [A Survey of Token Compression for Efficient Multimodal Large Language Models](https://mlsanity.com/paper/2507.20198/) (2025)
- [Efficient Multi-modal Large Language Models via Progressive Consistency Distillation](https://mlsanity.com/paper/2510.00515/) (2025)
- [Why 1 + 1 < 1 in Visual Token Pruning: Beyond Naive Integration via Multi-Objective Balanced Covering](https://mlsanity.com/paper/2505.10118/) (2025)
- [ViCA: Efficient Multimodal LLMs with Vision-Only Cross-Attention](https://mlsanity.com/paper/2602.07574/) (2026)
- [CLIP-MG: Guiding Semantic Attention with Skeletal Pose Features and RGB Data for Micro-Gesture Recognition on the iMiGUE Dataset](https://mlsanity.com/paper/2506.16385/) (2025)
- [Attention Debiasing for Token Pruning in Vision Language Models](https://mlsanity.com/paper/2508.17807/) (2025)
- [Revisiting MLLM Token Technology through the Lens of Classical Visual Coding](https://mlsanity.com/paper/2508.13460/) (2025)
- [Keep What Audio Cannot Say: Context-Preserving Token Pruning for Omni-LLMs](https://mlsanity.com/paper/2605.11605/) (2026)
- [OmniRefine: Alignment-Aware Cooperative Compression for Efficient Omnimodal Large Language Models](https://mlsanity.com/paper/2605.12056/) (2026)
- [Task-Related Token Compression in Multimodal Large Language Models from an Explainability Perspective](https://mlsanity.com/paper/2506.01097/) (2025)
- [Compression Tells Intelligence: Visual Coding, Visual Token Technology, and the Unification](https://mlsanity.com/paper/2601.20742/) (2026)
- [SVAC: Scaling Is All You Need For Referring Video Object Segmentation](https://mlsanity.com/paper/2509.24109/) (2025)
- [Spectral-Progressive Thought Flow for Lightweight Multimodal Reasoning](https://mlsanity.com/paper/2606.02842/) (2026)
- [DCP-Prune: Ultra-Low Token Pruning with Distribution Consistency Preservation](https://mlsanity.com/paper/2606.16633/) (2026)
- [Buried in Textual Debt: Context Pruning with Visual Evidence Preservation for MLLM Agents](https://mlsanity.com/paper/2608.22963/) (2026)
- [VisCache: Visual KV Cache Pruning for Efficient Vision Large Language Model Inference](https://mlsanity.com/paper/2608.24063/) (2026)
- [Doc-REFRAG: Rethinking Multimodal Document Retrieval-Augmented Generation](https://mlsanity.com/paper/2608.30163/) (2026)
- [DecoMoE: Decoupling Visual Propagation and Expert Computation for Efficient Multimodal MoE Inference](https://mlsanity.com/paper/2609.38823/) (2026)

## Abstract (arXiv)

Multimodal Large Language Models (MLLMs) are becoming increasingly popular, while the high computational cost associated with multimodal data input, particularly from visual tokens, poses a significant challenge. Existing training-based token compression methods improve inference efficiency but require costly retraining, while training-free methods struggle to maintain performance when aggressively reducing token counts. In this study, we reveal that the performance degradation of MLLM closely correlates with the accelerated loss of information in the attention output matrix. This insight introduces a novel information-preserving perspective, making it possible to maintain performance even under extreme token compression. Based on this finding, we propose TokenCarve, a training-free, plug-and-play, two-stage token compression framework. The first stage employs an Information-Preservation-Guided Selection (IPGS) strategy to prune low-information tokens, while the second stage further leverages IPGS to guide token merging, minimizing information loss. Extensive experiments on 11 datasets and 2 model variants demonstrate the effectiveness of TokenCarve. It can even reduce the number of visual tokens to 22.2% of the original count, achieving a 1.23x speedup in inference, a 64% reduction in KV cache storage, and only a 1.54% drop in accuracy. Our code is available at https://github.com/ShawnTan86/TokenCarve.