# ToFu: Visual Tokens Reduction via Fusion for Multi-modal, Multi-patch, Multi-image Task

arXiv: https://arxiv.org/abs/2503.04444 · HTML: https://mlsanity.com/paper/2503.04444/

## Key facts

- Cited by 3 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 33 of its 38 references that are in the map.
- First posted on arXiv 6 Mar 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Visual token pruning and efficiency](https://mlsanity.com/mllm/visual-token-pruning-and-efficiency/)

## Builds on (most cited first, 30 of 33)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [MMBench: Is Your Multi-modal Model an All-around Player?](https://mlsanity.com/paper/2307.06281/) (2023)
- [Phi-3 Technical Report: A Highly Capable Language Model Locally on Your Phone](https://mlsanity.com/paper/2404.14219/) (2024)
- [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)
- [How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suites](https://mlsanity.com/paper/2404.16821/) (2024)
- [Token Merging: Your ViT But Faster](https://mlsanity.com/paper/2210.09461/) (2022)
- [VideoChat: Chat-Centric Video Understanding](https://mlsanity.com/paper/2305.06355/) (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)
- [NExT-GPT: Any-to-Any Multimodal LLM](https://mlsanity.com/paper/2309.05519/) (2023)
- [LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models](https://mlsanity.com/paper/2407.07895/) (2024)
- [LLaVA-PruMerge: Adaptive Token Reduction for Efficient Large Multimodal Models](https://mlsanity.com/paper/2403.15388/) (2024)
- [ABO: Dataset and Benchmarks for Real-World 3D Object Understanding](https://mlsanity.com/paper/2110.06199/) (2021)
- [The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits](https://mlsanity.com/paper/2402.17764/) (2024)
- [mPLUG-Owl3: Towards Long Image-Sequence Understanding in Multi-Modal Large Language Models](https://mlsanity.com/paper/2408.04840/) (2024)
- [MM1: Methods, Analysis & Insights from Multimodal LLM Pre-training](https://mlsanity.com/paper/2403.09611/) (2024)
- [SlideVQA: A Dataset for Document Visual Question Answering on Multiple Images](https://mlsanity.com/paper/2301.04883/) (2023)
- [Pixtral 12B](https://mlsanity.com/paper/2410.07073/) (2024)
- [Revisiting the "Video" in Video-Language Understanding](https://mlsanity.com/paper/2206.01720/) (2022)
- [TinyLLaVA: A Framework of Small-scale Large Multimodal Models](https://mlsanity.com/paper/2402.14289/) (2024)
- [Matryoshka Multimodal Models](https://mlsanity.com/paper/2405.17430/) (2024)
- [TinyGPT-V: Efficient Multimodal Large Language Model via Small Backbones](https://mlsanity.com/paper/2312.16862/) (2023)
- [3D Concept Learning and Reasoning from Multi-View Images](https://mlsanity.com/paper/2303.11327/) (2023)
- [Adaptive Sparse ViT: Towards Learnable Adaptive Token Pruning by Fully Exploiting Self-Attention](https://mlsanity.com/paper/2209.13802/) (2022)
- [HiRED: Attention-Guided Token Dropping for Efficient Inference of High-Resolution Vision-Language Models](https://mlsanity.com/paper/2408.10945/) (2024)
- [Visual Question Answering on Image Sets](https://mlsanity.com/paper/2008.11976/) (2020)

## Built on by (most cited first, 3 of 3)

- [PARCEL: Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding](https://mlsanity.com/paper/2605.30126/) (2026)
- [From Saliency to Discriminability: Rank-Preserving Visual Token Pruning for VLM Rerankers](https://mlsanity.com/paper/2609.00667/) (2026)
- [Visual Tokens on Demand: Visual Budgeting for Scalable and Efficient Multi-Image Reasoning in Multimodal LLMs](https://mlsanity.com/paper/2610.18858/) (2026)

## Abstract (arXiv)

Large Multimodal Models (LMMs) are powerful tools that are capable of reasoning and understanding multimodal information beyond text and language. Despite their entrenched impact, the development of LMMs is hindered by the higher computational requirements compared to their unimodal counterparts. One of the main causes of this is the large amount of tokens needed to encode the visual input, which is especially evident for multi-image multimodal tasks. Recent approaches to reduce visual tokens depend on the visual encoder architecture, require fine-tuning the LLM to maintain the performance, and only consider single-image scenarios. To address these limitations, we propose ToFu, a visual encoder-agnostic, training-free Token Fusion strategy that combines redundant visual tokens of LMMs for high-resolution, multi-image, tasks. The core intuition behind our method is straightforward yet effective: preserve distinctive tokens while combining similar ones. We achieve this by sequentially examining visual tokens and deciding whether to merge them with others or keep them as separate entities. We validate our approach on the well-established LLaVA-Interleave Bench, which covers challenging multi-image tasks. In addition, we push to the extreme our method by testing it on a newly-created benchmark, ComPairs, focused on multi-image comparisons where a larger amount of images and visual tokens are inputted to the LMMs. Our extensive analysis, considering several LMM architectures, demonstrates the benefits of our approach both in terms of efficiency and performance gain.