AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Unveiling the Potential of Quantization with MXFP4: Strategies for Quantization Error Reduction

Jatin Chhugani, Geonhwa Jeong, Bor-Yiing Su, Yunjie Pan, Hanmei Yang, Aayush Ankit, Jiecao Yu, Summer Deng, Yunqing Chen, Nadathur Satish, Changkyu Kim

ICML 2026 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization · 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
add Overflow-Aware Scaling to block-16 MXFP4 (ablation)MXFP4-16 on Llama-3.1-8B-Instruct · Downstream averageaccuracy: 60.55 -> 63.39Improved
add Overflow-Aware Scaling to block-16 MXFP4 (ablation)MXFP4-16 on Qwen3-8B · Downstream averageaccuracy: 66.82 -> 69.71Improved
add static Macro Block Scaling on top of OAS (ablation)MXFP4-16-OAS on Llama-3.1-8B-Instruct · Downstream averageaccuracy: 63.39 -> 65.77Improved
add static Macro Block Scaling on top of OAS (ablation)MXFP4-16-OAS on Qwen3-8B · Downstream averageaccuracy: 69.71 -> 70.50Improved
use dynamic MBS for weights (hybrid) instead of static MBS everywhere (ablation)MXFP4-MBS-S on Llama-3.1-8B-Instruct · Downstream averageaccuracy: 65.77 -> 66.50Improved
use dynamic MBS for weights (hybrid) instead of static MBS everywhere (ablation)MXFP4-MBS-S on Qwen3-8B · Downstream averageaccuracy: 70.50 -> 70.84Improved
reduce block size from 32 to 16 alone (ablation)MXFP4-OCP on Llama-3.1-8B-Instruct · Downstream averageaccuracy: 61.25 -> 60.55Hurt
reduce block size from 32 to 16 alone (ablation)MXFP4-OCP on Qwen3-8B · Downstream averageaccuracy: 65.50 -> 66.82Improved
add OAS to block-16 MXFP4 (ablation)MXFP4-16 on DeepSeek-R1 · MMLU-Proaccuracy: 76.29 -> 75.36Hurt
add OAS to block-16 MXFP4 (ablation)MXFP4-16 on Llama-3.1-8B-Instruct · Wikitext perplexityperplexity: 11.52 -> 10.57Improved
replace OCP MXFP4 with MBS-HMXFP4-OCP on Qwen3-8B · MMLU-Proaccuracy: 43.78 -> 59.30Improved
increase MBS block size from 32 to 512 (ablation)MBS on Llama-3.1-8B-Instruct · Output QSNRQSNR (dB): approximately 1.1 dB degradationHurt

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