AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Activation Quantization of Vision Encoders Needs Prefixing Registers

Seunghyeon Kim, Taesun Yeom, Jinho Kim, Wonpyo Park, Kyuyeun Kim, Jaeho Lee

ECCV 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add RegCache on top of RepQ-ViT (CLIP-B/16)RepQ-ViT · ImageNet-1k zero-shot W4A4accuracy: 1.83 -> 21.50Improved
add RegCache on top of FIMA-Q (CLIP-B/16)FIMA-Q · ImageNet-1k zero-shot W4A4accuracy: 50.41 -> 62.08Improved
add RegCache on top of ERQ (SigLIP2-B/16)ERQ · ImageNet-1k zero-shot W4A4accuracy: 10.56 -> 28.98Improved
add RegCache on naive W8A8 quantization (SigLIP2)Naive RTN · ImageNet-1k zero-shot W8A8accuracy: 26.04 -> 72.35Improved
add RegCache to DINOv2 with advanced PTQ at W8A8PTQ4ViT/RepQ-ViT/ERQ · ImageNet-1k zero-shot W8A8 (DINOv2)accuracy: average -0.47Hurt
add RegCache at high bitwidth with strong PTQPTQ baselines · ImageNet-1k zero-shot W8A8accuracy: average +0.25 to +1.36No clear effect
add prefix caching only (no token deletion) (ablation)RegCache · ImageNet-1k W8A8 (SigLIP2)accuracy: 26.04 -> 23.82Hurt
add token deleting only (no prefix caching) (ablation)RegCache · ImageNet-1k W8A8 (SigLIP)accuracy: 69.71 -> 42.41Hurt
combine prefix caching with token deleting (ablation)RegCache · ImageNet-1k W8A8 (SigLIP2)accuracy: 26.04 -> 72.35Improved
add RegCache to RepQ-ViT in 4-bit VLM vision encoderQwen3-VL-2B · VQAv2accuracy: 32.03 -> 42.72Improved
add RegCache to RepQ-ViT in 4-bit VLM vision encoderQwen3-VL-8B · MLVUaccuracy: 44.89 -> 45.40Improved
add RegCache with ImageNet-searched prefix to other datasetsCLIP-B/16 W4A4 naive · StanfordCarsaccuracy: 29.76 -> 49.96Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy