AI Paper Atlas

Research map / Architectures and efficiency / Quantization

SSDi8: Accurate and Efficient 8-bit Quantization for State Space Duality

Hyunwoo Kim, Byoungchan Ko, Minseok Kang, Minwoo Kim, Dongjin Lee, Jaehoon Lee, Sungroh Yoon, Dahuin Jung

ICLR 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of ViTs · 0 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
Use SSDi8 instead of Quamba2 at W4A8Mamba-2 2.7B · six zero-shot tasks avgaccuracy: 62.1% -> 62.7%Improved
Use SSDi8 instead of Quamba2 at W8A8Mamba-2 8B · six zero-shot tasks avgaccuracy: 69.8% -> 70.2%Improved
Add SSD quantization to HadMamba (ablation)Mamba-2 2.7B W4A8 · LAMBADAaccuracy: 51.2% -> 67.2%Improved
Add mean correction (ablation)SSDi8 W4A8, Mamba-2 2.7B · LAMBADAaccuracy: 67.2% -> 67.4%Improved
Quantize B and C once and reuse (ablation)SSD W4A8 with ChunkState Q(X) · SSD latencylatency (ms): 8.58 -> 8.05Improved
Add ChunkBMM quantization on top of reformulation and persistent INT8 (ablation)SSDi8 W4A8 · SSD latencylatency (ms): 8.00 -> 6.53Improved
Use sparse-aware reformulation instead of quantizing X alone (ablation)SSDi8 W4A8 · WikiText2perplexity: 9.35 -> 9.36No clear effect
Quantize X and State per head (H,G) together with P instead of per-T (ablation)SSD W4A8 · WikiText2perplexity: 11.97 -> 7.42Improved
Naively quantize SSD per-tensor after in-projection (ablation)Mamba-2 2.7B W4A8 · six zero-shot tasks avgaccuracy: 63.6% -> 58.4%Hurt
Apply INT8 to the SSD path onlyNemotron-H-8B-Reasoning · SSD module latencylatency (ms): 19.834 -> 9.156Improved
Apply INT8 to the SSD path onlyNemotron-H-8B-Reasoning · average zero-shot accuracyaccuracy: 73.1% -> 73.0%No clear effect
Use SSDi8 instead of Quamba2 on edge deviceMamba-2 2.7B W8A8 · Orin NX SSD latency L=2048latency (ms): 249.29 -> 217.69Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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