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.
arXiv:2610.03389v1 Announce Type: cross Abstract: Vision language models (VLMs) incur substantial inference cost because every visual token is processed by the attention and MLP projections of every...
arXiv:2508.03351v3 Announce Type: replace-cross Abstract: Large language models (LLMs) have demonstrated remarkable capabilities across diverse language tasks, motivating their extension to vision-la...
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.
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.
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...
VisionWeave introduces elastic visual representation weaving, a native capability for multimodal large language models that learns where and at what granularity to encode visual information. The method combines a gated spatial pooler for coarse representations with a granularity router that allocates content‑adaptive token usage, trained end‑to‑end on large‑scale data. Experiments on Qwen3.5‑4B and Qwen3.8‑27B show that VisionWeave can save 43.0% of tokens while preserving 98.9% of performance across eight benchmarks, and delivers significant throughput gains and latency reductions when deployed on the SGLang serving engine.
arXiv:2606. 11576v1 Announce Type: cross Abstract: Modern Vision-Language Models (VLMs) benefit from chain-of-thought prompting and test-time scaling, but these gains often come with prohibitive inference cost due to large visual contexts and long decoding chains.
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:2607. 25527v1 Announce Type: cross Abstract: Unifying visual understanding and generation in one model holds immense promise, but remains challenging and expensive due to heavy compute and data demands and conflicts between the visual features needed for these two capabilities.
arXiv:2609.16722v1 Announce Type: new Abstract: Scaling Multimodal Large Language Models (MLLMs) to long-form video understanding is bottlenecked by the explosion of visual tokens, which saturates co...
LensVLM is an inference framework and post‑training recipe that lets Vision‑Language Models (VLMs) process compressed images of text by selectively expanding only the relevant parts back to full resolution. Using Qwen3.5‑9B‑Base, LensVLM achieves accuracy comparable to full‑text models at 4.3× compression and outperforms other compression baselines up to 10.1× across seven text QA benchmarks, while also improving performance on multimodal document and code tasks as compression increases.
arXiv:2606. 01503v1 Announce Type: cross Abstract: Unified vision-language models (VLMs) integrate visual understanding and visual generation within a single autoregressive backbone, but their joint training is computationally expensive and largely overlooked from an efficiency perspective.