mlsanityData to 14 Oct 2026

Channel-Wise and Token-Aware Post-Training Quantization for Visual State Space Duality

arXiv 2609.16656 ↗First posted 15 Sep 2026cs.CVcited by 0

On the map: State-space sequence models (S4, Mamba) → Vision Mamba backbones

Abstract (arXiv)

State space models (SSMs), particularly Mamba, have emerged as efficient alternatives to attention-based architectures and have been extended to vision through ViM, VMamba, and Visual State Space Duality (VSSD). Yet the low-bit post-training quantization (PTQ) behavior of VSSD remains insufficiently understood. A weight-activation split on VSSD-Tiny identifies activation quantization as the dominant low-bit bottleneck, while representative inputs to selected VSSD-backbone linear layers exhibit strong channel-wise magnitude variation and token-localized extremes. We propose the Channel-wise Token-balanced Output-Aware Clipping (CTOAC) method, which learns per-input-channel clipping bounds by minimizing a token-balanced reconstruction loss on the corresponding linear outputs. Only the selected linear layers and their input activations are quantized; other backbone operations retain their original precision. Across VSSD-Tiny, VSSD-Small, and VSSD-Base, the proposed CTOAC method retains ImageNet-1K accuracy and remains substantially more robust than the evaluated baselines at more aggressive precision settings. Applying the same quantization scope to VSSD backbones on COCO and ADE20K preserves strong object detection, instance segmentation, and semantic segmentation performance. An optimized RTX 4090 deployment configuration achieves up to 1.42x end-to-end speedup over FP32.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Vision Mamba backbones, one of 10 directions in State-space sequence models (S4, Mamba). Builds on 35 of its 40 references that are in the map. First posted on arXiv 15 Sep 2026.

Builds on

35 of 40 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025OuroMamba: A Data-Free Quantization Framework for Vision Mamba13 Mar
  2. 2025ViM-VQ: Efficient Post-Training Vector Quantization for Visual Mamba12 Mar
  3. 2025Post-Training Quantization for Vision Mamba with k-Scaled Quantization and Reparameterization28 Jan
  4. 2024PTQ4VM: Post-Training Quantization for Visual Mamba29 Dec
  5. 2024VSSD: Vision Mamba with Non-Causal State Space Duality26 Jul
  6. 2024LocalMamba: Visual State Space Model with Windowed Selective Scan14 Mar
  7. 2024Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model17 Jan

Other references · 28

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2021Swin Transformer: Hierarchical Vision Transformer using Shifted Windowscited by 12,139
  4. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,498
  5. 2023Mamba: Linear-Time Sequence Modeling with Selective State Spacescited by 4,718
  6. 2021Efficiently Modeling Long Sequences with Structured State Spacescited by 2,465
  7. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,079
  8. 2017Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inferencecited by 1,593
  9. 2022SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Modelscited by 1,386
  10. 2024Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Dualitycited by 1,336
  11. 2024VMamba: Visual State Space Modelcited by 969
  12. 2020Up or Down? Adaptive Rounding for Post-Training Quantizationcited by 621
16 more
  1. 2021BRECQ: Pushing the Limit of Post-Training Quantization by Block Reconstructioncited by 479
  2. 2021Post-Training Quantization for Vision Transformercited by 278
  3. 2021PTQ4ViT: Post-training quantization for vision transformers with twin uniform quantizationcited by 179
  4. 2021FQ-ViT: Post-Training Quantization for Fully Quantized Vision Transformercited by 151
  5. 2022RepQ-ViT: Scale Reparameterization for Post-Training Quantization of Vision Transformerscited by 137
  6. 2023Towards Accurate Post-Training Quantization for Vision Transformercited by 71
  7. 2022NoisyQuant: Noisy Bias-Enhanced Post-Training Activation Quantization for Vision Transformerscited by 61
  8. 2024AdaLog: Post-Training Quantization for Vision Transformers with Adaptive Logarithm Quantizercited by 29
  9. 2025MambaQuant: Quantizing the Mamba Family with Variance Aligned Rotation Methodscited by 29
  10. 2024Quamba: A Post-Training Quantization Recipe for Selective State Space Modelscited by 28
  11. 2024Instance-Aware Group Quantization for Vision Transformerscited by 25
  12. 2024DopQ-ViT: Towards Distribution-Friendly and Outlier-Aware Post-Training Quantization for Vision Transformerscited by 19
  13. 2025APHQ-ViT: Post-Training Quantization with Average Perturbation Hessian Based Reconstruction for Vision Transformerscited by 19
  14. 2025FIMA-Q: Post-Training Quantization for Vision Transformers by Fisher Information Matrix Approximationcited by 19
  15. 2025QMamba: Post-Training Quantization for Vision State Space Modelscited by 4
  16. 2026SSDi8: Accurate and Efficient 8-bit Quantization for State Space Dualitycited by 2

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026SCULPT: Training Edge Vision Models for Post-Training Quantization Readiness
  2. 2026The Rescue Effect: Spatio-Semantic Early Exit Bypasses Quantization Collapse in CLIP
  3. 2025TernaryCLIP: Efficiently Compressing Vision-Language Models with Ternary Weights and Distilled Knowledge
  4. 2024V"Mean"ba: Visual State Space Models only need 1 hidden dimension
  5. 2026A Unified Rate-Distortion Perspective on Vector, Product, and Scalar Quantization
  6. 2023I&S-ViT: An Inclusive & Stable Method for Pushing the Limit of Post-Training ViTs Quantization
  7. 2026Stateful Token Reduction for Long-Video Hybrid VLMs
  8. 2026Ternary Mamba: Grouped Quantization-Aware Training of W1.58A16 State Space Models