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...
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.
In video understanding, vision-language models (VLMs) must ingest massive numbers of visual tokens, causing the computational and memory cost of the prefill stage to rise sharply. Such visual sequences are highly redundant along the spatio-temporal dimension, yet a high compression ratio is often accompanied by the loss of critical details.
arXiv:2512. 08240v2 Announce Type: replace-cross Abstract: Vision-language models (VLMs) rely on hundreds of visual tokens, leading to high computational and memory costs.
The paper introduces Varest, a method for image-token compression that reduces computational load while preserving semantic integrity. Varest uses variable-resolution tokenization guided by semantic fungibility, merging tokens within coherent regions coarsely and maintaining finer distinctions at object boundaries and text. After early decoder processing, learned relevance registers prune visual tokens, enabling a single checkpoint to perform well across multiple budgets and backbones, achieving 86.10‑92.33% of full-token performance with only 5.56% of the original token count.
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...
The paper introduces ReFIT, an instruction‑guided visual token reduction framework designed to accelerate large vision‑language model inference. ReFIT combines Relevance‑Guided Window Reshaping (RWR) to adaptively capture instruction‑relevant regions and Instruction‑Guided Token Refinement (ITR) to prune unnecessary visual tokens. Experiments on four VQA benchmarks show that ReFIT improves answer accuracy while lowering computational cost, and qualitative results confirm its ability to localize relevant regions and remove extraneous visual information.
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...
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:2609.35232v2 Announce Type: replace-cross Abstract: Visual-token compression is effective for improving the efficiency of vision-language models, but under extreme compression budgets, token pr...
Vision-language models commonly project all tokens produced by a pretrained vision encoder into a large language model. However, final-layer features can discard text, local attributes, and spatial relationships, while high-resolution inputs substantially increase context length and inference latency.
arXiv:2609.39924v1 Announce Type: cross Abstract: Vision-language models face a fundamental scaling bottleneck: the number of visual tokens grows with both temporal duration and spatial resolution, m...