AI Paper Atlas

Research map / Architectures and efficiency / Quantization

MixA-Q: Revisiting Activation Sparsity for Vision Transformers from a Mixed-Precision Quantization Perspective

Weitian Wang, Rai Shubham, Cecilia De La Parra, Akash Kumar

ICCV 2025 · Architectures and efficiency › Quantization · approach: Post-training quantization of ViTs · 4 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 sparsity-aware quantization adaptation (MixA-Q at full 4-bit) (ablation)OFQ W4A4 Mask R-CNN Swin-T · COCO detmAP: 43.1 -> 43.8Improved
use MixA-Q instead of uniform 3-bit activationsOFQ W4A4 · COCO det, 1.35xmAP: 41.4 -> 42.3Improved
use MixA-Q instead of SparseViT at 1.24xOFQ W4A4 · COCO det, 1.24xmAP: 43.1 -> 43.2No clear effect
add 30% activation pruning to MixA-QMixA-Q · COCO det, 1.53xmAP: 41.4 -> 42.1Improved
add 30% activation pruning to MixA-QMixA-Q · COCO det, 1.82xmAP: 40.1 -> 40.5Improved
use uniform-sum compression ratio sampling instead of naive uniform sampling (ablation)SAQA · COCO det, 3.36 bitmAP: 43.0 -> 43.4Improved
use uniform-sum compression ratio sampling instead of naive uniform sampling (ablation)SAQA · COCO det, 3 bitmAP: 41.9 -> 42.4Improved
start SAQA from the W4A4-quantized model rather than floating point (ablation)SAQA · COCO det, 3.36 bitmAP: 42.8 -> 43.4Improved
allow elevating compressed windows back to high precision in later blocks (ablation)MixA-Q · COCO detmAP: -0.1%Hurt
use MixA-Q instead of SparseViT on out-of-distribution weather imagesOFQ W4A4 · COCO-OmAP_o: 29.6 -> 30.5Improved
use MixA-Q instead of SparseViT under worst window selectionOFQ W4A4 · COCO worst window selectionmAP_w: 24.0 -> 28.8Improved
use MixA-Q instead of SparseViT on clean COCO at 1.4xOFQ W4A4 · COCO det, 1.4xmAP: 42.4 -> 41.5Hurt

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