The paper introduces Adaptive Visual Token Pruning (AVTP), a training‑free framework that dynamically selects pruning layers and ratios for large vision‑language models (LVLMs) when processing multiple image sequences. By analyzing visual attention distributions across different LVLM architectures, AVTP adapts token retention to image importance, enabling efficient inference without relying on attention‑based computations incompatible with FlashAttention. Experiments show significant speedups—up to 2× for Qwen3VL‑8B—while preserving or even improving accuracy on multi‑image benchmarks.
arXiv:2609.05916v1 Announce Type: cross Abstract: Large vision-language models (LVLMs) achieve strong multimodal understanding, but the hundreds to thousands of visual tokens they process impose subs...
The paper shows that only a small subset of attention heads in vision-language models is responsible for selecting critical visual tokens. By pruning tokens based on similarity before LLM reasoning and then applying head‑aware pruning during reasoning, the proposed ProViP framework achieves high task performance with significant speedups. Experiments on LLaVA‑1.5‑7B demonstrate 95.9% performance retention and a 1.62× inference speedup at an 88.9% pruning ratio.
arXiv:2609.10346v1 Announce Type: new Abstract: Multimodal large language models (MLLMs) process hundreds or thousands of visual tokens per image, incurring prohibitive inference costs. While existin...
The paper introduces V-CoLA, a training‑free token compression framework tailored for vision-language models that use linear attention. V-CoLA employs a uniqueness‑aware importance criterion to select critical vision tokens and an adaptive token merging strategy to compress them, all optimized to fit the chunk‑wise parallelism of linear attention. Experiments show that V‑CoLA retains 99.5% of the original performance with only half the vision tokens and achieves significant speedups during prefill.
arXiv:2609.24485v1 Announce Type: new Abstract: Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction oft...
arXiv:2608. 03112v1 Announce Type: cross Abstract: Vision-language models excel at image and video understanding but suffer from high inference latency due to the need to process thousands of tokens per image, limiting their deployment on resource-constrained edge devices and in real-time surveillance applications.
LearnPruner rethinks token pruning in vision‑language models by addressing shortcomings of attention mechanisms in vision encoders and large language models. The method uses a two‑stage approach: first a learnable pruning module removes redundant vision tokens after the vision encoder, then only task‑relevant tokens are kept in the LLM’s middle layer. Experiments show that LearnPruner retains about 95% of performance while using only 5.5% of vision tokens, achieving a 3.2× speed‑up.
SinkPruner is a training‑free framework that prunes visual tokens for multimodal large language models by first removing high‑norm redundant tokens with a visual sanitizer and then selectively keeping tokens that align with the text query using a text‑guided pruner. The coarse‑to‑fine design reduces attention sink and dispersion, enabling an 89% token reduction while preserving 96.5% of LLaVA‑1.5’s performance and 91.8% of Qwen2.5‑VL’s performance across twelve image‑language and four video‑language benchmarks. The visual sanitizer also improves existing pruning methods, showing strong transferability.
PACE introduces a training‑free Condense‑and‑Extract framework that speeds up Vision‑Language Model inference by first adaptively downsampling visual inputs before encoding and then selectively retaining essential tokens during decoding. The Adaptive Pixel Compressor (APC) reduces encoder workload while preserving global context, and the Dynamic Dual‑Attention Extractor (DDAE) keeps task‑critical details by fusing visual and language signals. Applied to Qwen2.5‑VL‑7B, PACE maintains 93.8% of performance using only 10% of visual tokens, achieving a 3.1× speedup in time to first token.
The paper introduces STD, a hierarchical token pruning framework for Large Vision‑Language Models that aligns pruning strategies with the functional roles of different network stages. By using high‑frequency spectral analysis in shallow layers, Gaussian‑smoothed attention in intermediate layers, and a stability‑adaptive trigger in deep layers, STD preserves essential visual information while aggressively reducing token counts. Experiments demonstrate that STD outperforms existing pruning methods, achieving up to 94.4% token reduction and a 3.9× speed‑up on LLaVA‑NeXT‑7B.
Visual token pruning is a promising approach to reducing the inference cost of large vision-language models (LVLMs), yet aggressive token reduction often causes substantial performance degradation. We...
ERASE is an adaptive two-stage token pruning framework designed to reduce computational overhead in Vision‑Language Models by eliminating redundant visual tokens. Stage 1 removes image‑level redundancy using lightweight raw‑image statistics, while Stage 2 progressively prunes instruction‑irrelevant tokens across decoder layers. Experiments on Qwen2.5‑VL‑7B show that ERASE retains 95.70 % of the original accuracy while keeping only 25 % of the tokens.