token-pruning

[EMNLP 2024 & AAAI 2026] A powerful toolkit for compressing large models including LLMs, VLMs, and video generative models.

This is a collection of our research on efficient AI, covering hardware-aware NAS and model compression.

[CVPR 2025] DivPrune: Diversity-based Visual Token Pruning for Large Multimodal Models

[NeurIPS 2025] HoliTom: Holistic Token Merging for Fast Video Large Language Models

[CVPR 2025] PACT: Pruning and Clustering-Based Token Reduction for Faster Visual Language Models

Official implementation of CVPR 2024 paper "vid-TLDR: Training Free Token merging for Light-weight Video Transformer".

[ICLR 2026] The official repo of "MMTok: Multimodal Coverage Maximization for Efficient Inference of VLMs"

Official Implementation (Pytorch) of the "Representation Shift: Unifying Token Compression with FlashAttention", ICCV 2025

Implementation of ICCV 2025 paper "Growing a Twig to Accelerate Large Vision-Language Models".

A token pruning method that accelerates ViTs for various tasks while maintaining high performance.

[ICLR 2026] AgilePruner: An Empirical Study of Attention and Diversity for Adaptive Visual Token Pruning in Large Vision-Language Models

[ECCV 2026] 3DZip: Spatial-Aware Feature Diversity-Guided Token Compression for 3D Question Answering

An implementation of LazyLLM token pruning for LLaMa 2 model family.

PriorTR (ECCV 2026): training-free, prior-corrected visual token reduction for accelerating multimodal LLMs — image & video.

Contract and claim-boundary tooling for empirical AI-system experiments. Turns retained evidence into falsifiable, condition-specific decisions.

Query-aware visual token pruning for VLMs. Five-component pipeline (cross-attention scorer, entropy controller, spatial coherence, progressive schedule, token recycling) attached via PyTorch forward hooks — no upstream model changes. Evaluated on POPE with LLaVA-1.5-7B.

Visual token budget simulator for efficient VLM inference

Token pruning and sparse attention for efficient Vision-Language Model inference

Index-Preserving Lightweight Token Pruning for Efficient Document Understanding in Vision-Language Models (ICLR 2026 Workshop on MM Intelligence, Poster).