AI Paper Atlas

Research map / Architectures and efficiency / Quantization

OuroMamba: A Data-Free Quantization Framework for Vision Mamba

Akshat Ramachandran, Mingyu Lee, Huan Xu, Souvik Kundu, Tushar Krishna

ICCV 2025 · Architectures and efficiency › Quantization · approach: Post-training quantization of ViTs · 6 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 patch-level contrastive loss L_C in place of patch-similarity loss L_PSE (ablation)OuroMamba-Gen (loss with L_O) · ImageNet-1K Vim-S W4A8top-1 accuracy: 73.45 -> 79.81Improved
add output loss L_O to contrastive loss (ablation)OuroMamba-Gen · ImageNet-1K Vim-S W4A8top-1 accuracy: 75.52 -> 79.81Improved
add output loss L_O to patch-similarity loss (ablation)OuroMamba-Gen · ImageNet-1K Vim-S W4A8top-1 accuracy: 71.68 -> 73.45Improved
add L_PSE on top of L_C + L_O (ablation)OuroMamba-Gen · ImageNet-1K Vim-S W4A8top-1 accuracy: 79.81 -> 74.32Hurt
use only output loss L_O for generation (ablation)OuroMamba-Gen · ImageNet-1K Vim-S W4A8top-1 accuracy: 21.65 vs 75.52 for L_C onlyHurt
calibrate with ViT-generated synthetic data instead of real dataQMamba · ImageNet-1K Vim-S W4A8top-1 accuracy: 7-14% accuracy dropHurt
use delta-based weighting instead of uniform weighting for patched state (ablation)OuroMamba-Gen · ImageNet-1K VMamba-B W4A4top-1 accuracyImproved
vary scanning direction used for generation (ablation)OuroMamba-Gen · ImageNet-1K VMamba-B W4A4top-1 accuracyNo clear effect
increase synthetic batch size beyond 128 (ablation)OuroMamba-Gen · ImageNet-1K Vim-S W4A8top-1 accuracyNo clear effect
use 5x5 neighborhood with top-12 positive patches (ablation)OuroMamba-Gen · ImageNet-1K Vim-S W4A8top-1 accuracyImproved
set periodic outlier list refresh to 10 time-steps (ablation)OuroMamba-Quant · W4A4 Vim-B and VMamba-Bnormalized speedupImproved
remove outlier list refresh (n_refresh = Full) (ablation)OuroMamba-Quant · W4A4 Vim-B and VMamba-BspeedupHurt

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