# Hybrid-Level Instruction Injection for Video Token Compression in Multi-modal Large Language Models

arXiv: https://arxiv.org/abs/2503.16036 · HTML: https://mlsanity.com/paper/2503.16036/

## 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 64 of its 75 references that are in the map.
- First posted on arXiv 20 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 64)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Flamingo: a Visual Language Model for Few-Shot Learning](https://mlsanity.com/paper/2204.14198/) (2022)
- [Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution](https://mlsanity.com/paper/2409.12191/) (2024)
- [Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context](https://mlsanity.com/paper/2403.05530/) (2024)
- [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)
- [LLaVA-OneVision: Easy Visual Task Transfer](https://mlsanity.com/paper/2408.03326/) (2024)
- [MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Models](https://mlsanity.com/paper/2304.10592/) (2023)
- [CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer](https://mlsanity.com/paper/2408.06072/) (2024)
- [Ego4D: Around the World in 3,000 Hours of Egocentric Video](https://mlsanity.com/paper/2110.07058/) (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)
- [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)
- [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)
- [VideoChat: Chat-Centric Video Understanding](https://mlsanity.com/paper/2305.06355/) (2023)
- [Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs](https://mlsanity.com/paper/2406.16860/) (2024)
- [VideoLLaMA 2: Advancing Spatial-Temporal Modeling and Audio Understanding in Video-LLMs](https://mlsanity.com/paper/2406.07476/) (2024)
- [NTU RGB+D: A Large Scale Dataset for 3D Human Activity Analysis](https://mlsanity.com/paper/1604.02808/) (2016)
- [EgoSchema: A Diagnostic Benchmark for Very Long-form Video Language Understanding](https://mlsanity.com/paper/2308.09126/) (2023)
- [LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models](https://mlsanity.com/paper/2407.07895/) (2024)
- [LLaVA-Video: Video Instruction Tuning With Synthetic Data](https://mlsanity.com/paper/2410.02713/) (2024)
- [Otter: A Multi-Modal Model with In-Context Instruction Tuning](https://mlsanity.com/paper/2305.03726/) (2023)

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

- [Token Reduction Should Go Beyond Efficiency in Generative Models -- From Vision, Language to Multimodality](https://mlsanity.com/paper/2505.18227/) (2025)
- [LLaVA-Scissor: Token Compression with Semantic Connected Components for Video LLMs](https://mlsanity.com/paper/2506.21862/) (2025)
- [See&Trek: Training-Free Spatial Prompting for Multimodal Large Language Model](https://mlsanity.com/paper/2509.16087/) (2025)
- [Beyond Hard and Soft: Hybrid Context Compression for Balancing Local and Global Information Retention](https://mlsanity.com/paper/2505.15774/) (2025)
- [EchoingPixels: Aliasing-Resistant Joint Token Reduction for Audio-Visual LLMs](https://mlsanity.com/paper/2512.10324/) (2025)
- [ShowTable: Unlocking Creative Table Visualization with Collaborative Reflection and Refinement](https://mlsanity.com/paper/2512.13303/) (2025)
- [Token Reduction via Local and Global Contexts Optimization for Efficient Video Large Language Models](https://mlsanity.com/paper/2603.01400/) (2026)
- [AIBench: Evaluating Visual-Logical Consistency in Academic Illustration Generation](https://mlsanity.com/paper/2603.28068/) (2026)
- [VCM: Vision Concept Modeling Based on Implicit Contrastive Learning with Vision-Language Instruction Fine-Tuning](https://mlsanity.com/paper/2504.19627/) (2025)
- [Enhancing Diffusion-based Unrestricted Adversarial Attacks via Adversary Preferences Alignment](https://mlsanity.com/paper/2506.01511/) (2025)
- [DynImg: Key Frames with Visual Prompts are Good Representation for Multi-Modal Video Understanding](https://mlsanity.com/paper/2507.15569/) (2025)
- [TransPrune: Token Transition Pruning for Efficient Large Vision-Language Model](https://mlsanity.com/paper/2507.20630/) (2025)
- [SToRM: Supervised Token Reduction for Multi-modal LLMs toward efficient end-to-end autonomous driving](https://mlsanity.com/paper/2602.11656/) (2026)
- [Exploiting Discriminative Codebook Prior for Autoregressive Image Generation](https://mlsanity.com/paper/2508.10719/) (2025)
- [VSI: Visual Subtitle Integration for Keyframe Selection to enhance Long Video Understanding](https://mlsanity.com/paper/2508.06869/) (2025)
- [VideoScaffold: Elastic-Scale Visual Hierarchies for Streaming Video Understanding in MLLMs](https://mlsanity.com/paper/2512.22226/) (2025)
- [GroundVTS: Visual Token Sampling in Multimodal Large Language Models for Video Temporal Grounding](https://mlsanity.com/paper/2604.02093/) (2026)
- [An Efficient Token Compression Framework for Visual Object Tracking](https://mlsanity.com/paper/2605.08329/) (2026)
- [Bridging Brain and Semantics: A Hierarchical Framework for Semantically Enhanced fMRI-to-Video Reconstruction](https://mlsanity.com/paper/2605.14569/) (2026)

## Abstract (arXiv)

Recent Multi-modal Large Language Models (MLLMs) have been challenged by the computational overhead resulting from massive video frames, often alleviated through compression strategies. However, the visual content is not equally contributed to user instructions, existing strategies (eg, average pool) inevitably lead to the loss of potentially useful information. To tackle this, we propose the Hybrid-level Instruction Injection Strategy for Conditional Token Compression in MLLMs (HICom), utilizing the instruction as a condition to guide the compression from both local and global levels. This encourages the compression to retain the maximum amount of user-focused information while reducing visual tokens to minimize computational burden. Specifically, the instruction condition is injected into the grouped visual tokens at the local level and the learnable tokens at the global level, and we conduct the attention mechanism to complete the conditional compression. From the hybrid-level compression, the instruction-relevant visual parts are highlighted while the temporal-spatial structure is also preserved for easier understanding of LLMs. To further unleash the potential of HICom, we introduce a new conditional pre-training stage with our proposed dataset HICom-248K. Experiments show that our HICom can obtain distinguished video understanding ability with fewer tokens, increasing the performance by 2.43% average on three multiple-choice QA benchmarks and saving 78.8% tokens compared with the SOTA method. The code is available at https://github.com/lntzm/HICom.