AI Paper Atlas

Research map / Architectures and efficiency / Quantization

BATQuant: Outlier-Resilient MXFP4 Quantization via Learnable Block-wise Optimization

Jifu Li, Manyi Zhang, Xiaobo Xia, Han Bao, Haoli Bai, Zhenhua Dong, Xianzhi Yu

ECCV 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 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
use block-wise affine transformation instead of global (ablation)BATQuant · Qwen3-8B W4A4KV16 non-reasoning avgaverage accuracy: 68.24 -> 68.70Improved
use block-wise affine transformation instead of global (ablation)BATQuant · Qwen3-VL-8B-Instruct W4A4KV16recovery rate (%): 95.59 -> 96.43Improved
add block-wise learnable clipping instead of global clipping (ablation)BATQuant · Qwen3-8B W4A4KV16 non-reasoning avgaverage accuracy: 68.51 -> 68.70Improved
add block-wise learnable clipping instead of global clipping (ablation)BATQuant · Qwen3-VL-8B-Instruct W4A4KV16recovery rate (%): 96.18 -> 96.43Improved
replace Naive Kronecker with Global and Private Kronecker (GPK) (ablation)BATQuant · parameter count (N=4096, g=32)parameters: 10,240 -> 2,112Improved
set affine block size equal to MXFP block size g=32 instead of smaller (16) or larger (128) (ablation)BATQuant · Qwen3-8B and Qwen3-VL-8BaccuracyImproved
vary global shared matrix size g1 (1 to 32) (ablation)BATQuant GPK · Qwen3-8B and Qwen3-VL-8BaccuracyMixed
replace FlatQuant with BATQuantQwen3-VL-8B-Instruct · W4A4KV16 multimodalrecovery rate (%): 94.79 -> 96.43Improved
replace GPTQ with BATQuantQwen3-8B · W4A8KV16 reasoningrecovery rate (%): 96.54 -> 97.46Improved
replace GPTQ with BATQuantQwen3-8B · W4A4KV16 reasoningrecovery rate (%): 90.10 -> 92.45Improved
replace GPTQ with BATQuantQwen3-8B · W4A8KV8 reasoningrecovery rate (%): 94.54 -> 96.22Improved
replace rotation-based QuaRot with BATQuantQwen3-VL-8B-Instruct · W4A4KV16 multimodalrecovery rate (%): 89.69 -> 96.43Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 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