SinkPruner: Sink-Free Visual Token Pruning for Multimodal Large Language Models

arXiv:2609.01004 · cs.CV, cs.AI, cs.CL, cs.LG · Submitted 2026-09-01 · Read on arXiv

cs.CV, cs.AI, cs.CL, cs.LG

Submitted: 2026-09-01

Updated: 2026-09-01

Code: https://github.com/LaVi-Lab/SinkPruner

Terminology

• Pruning — 119× in this paper · explained in 4 episodes
• Benchmark — 55× in this paper · explained in 2 episodes
• Redundancy — 17× in this paper · explained in 2 episodes
• Transformer — 14× in this paper · explained in 2 episodes
• Accuracy — 11× in this paper · explained in 1 episode
• Hyperparameters — 7× in this paper · explained in 2 episodes
• Transferability — 7× in this paper · explained in 4 episodes
• Hallucination — 6× in this paper · explained in 4 episodes
• Robustness — 6× in this paper · explained in 5 episodes
• Ablation — 4× in this paper · explained in 1 episode
• Aggregation — 4× in this paper · explained in 1 episode
• Generalizability — 4× in this paper · explained in 6 episodes
• Instruction Tuning — 4× in this paper · explained in 2 episodes
• Vision-language models — 4× in this paper · explained in 2 episodes
• Calibration — 3× in this paper · explained in 7 episodes
• Computer Vision and Pattern Recognition — 3× in this paper
• Cosine Similarity — 3× in this paper · explained in 3 episodes
• Fidelity — 3× in this paper · explained in 2 episodes
• Ablation Studies — 2× in this paper · explained in 1 episode
• Alignment — 2× in this paper · explained in 1 episode
• Attention Entropy — 2× in this paper · explained in 2 episodes
• Attention Mechanism — 2× in this paper · explained in 3 episodes
• Computational overhead — 2× in this paper · explained in 2 episodes
• Core — 2× in this paper · explained in 2 episodes
• Embeddings — 2× in this paper · explained in 2 episodes
• Generalization — 2× in this paper · explained in 13 episodes
• Multimodal Reasoning — 2× in this paper · explained in 5 episodes
• Sparse Attention — 2× in this paper · explained in 5 episodes
• Temporal Reasoning — 2× in this paper · explained in 2 episodes
• Abstraction — 1× in this paper · explained in 1 episode
• Adaptability — 1× in this paper · explained in 1 episode
• Benchmarking — 1× in this paper · explained in 4 episodes
• Clustering — 1× in this paper · explained in 2 episodes
• Fine-Tuning — 1× in this paper · explained in 5 episodes
• Grounding — 1× in this paper · explained in 3 episodes
• LLM judge — 1× in this paper · explained in 2 episodes
• Prompt Engineering — 1× in this paper · explained in 4 episodes
• Representation Learning — 1× in this paper · explained in 2 episodes
• Semantic Alignment — 1× in this paper · explained in 3 episodes
• Tokenization — 1× in this paper · explained in 2 episodes
• Transformer Architecture — 1× in this paper · explained in 2 episodes

Sources

• Qwen Technical Report
• Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond
• Qwen3-VL Technical Report
• Qwen2.5-VL Technical Report
• Perception Encoder: The best visual embeddings are not at the output of the network
• MMTok: Multimodal Coverage Maximization for Efficient Inference of VLMs
• MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Models
• Mini-Gemini: Mining the Potential of Multi-modality Vision Language Models
• ApET: Approximation-Error Guided Token Compression for Efficient VLMs
• RoboMamba: Efficient Vision-Language-Action Model for Robotic Reasoning and Manipulation
• Multi-Stage Vision Token Dropping: Towards Efficient Multimodal Large Language Model
• Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context
• Gemini: A Family of Highly Capable Multimodal Models
• Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
• LLaMA: Open and Efficient Foundation Language Models
• AutoPrune: Each Complexity Deserves a Pruning Policy
• InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
• HiDrop: Hierarchical Vision Token Reduction in MLLMs via Late Injection, Concave Pyramid Pruning, and Early Exit
• PyramidDrop: Accelerating Your Large Vision-Language Models via Pyramid Visual Redundancy Reduction
• Qwen3 Technical Report

Related papers

• Loss Knows Best: Detecting Annotation Errors in Videos via Loss Trajectories
• AnchorWeave: World-Consistent Video Generation with Retrieved Local Spatial Memories
• Benchmarking the Robustness of Foundation Models for Mammography under Domain Shift
• MambaX-Net: Dual-Input Mamba-Enhanced Cross-Attention Network for Longitudinal MRI Segmentation
• TeleOCR: Navigating Document Parsing Across Digital and Camera-Captured Documents
• A Survey on Efficient Vision-Language-Action Models