AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Bridging the Gap Between Promise and Performance for FP4 Quantization

Vage Egiazarian, Roberto L. Castro, Denis Kuznedelev, Andrei Panferov, Eldar Kurtic, Shubhra Pandit, Alexandre Noll Marques, Mark Kurtz, Saleh Ashkboos, Torsten Hoefler, Dan Alistarh

ICLR 2026 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization

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 Hadamard rotation (group size) to RTN (ablation)NVFP4 RTN · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 74.73 -> 74.05Hurt
add Hadamard rotation to RTN (ablation)MXFP4 RTN · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 69.32 -> 70.45Improved
replace RTN with MR-GPTQMXFP4 W4A4 · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 69.32 -> 73.65Improved
replace GPTQ with MR-GPTQMXFP4 W4A4 · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 70.62 -> 73.65Improved
replace GPTQ with MR-GPTQNVFP4 W4A4 · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 75.72 -> 75.84No clear effect
add QuaRot rotation to RTNMXFP4 W4A4 · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 69.32 -> 62.90Hurt
add Hadamard to RTN (ablation)INT4 W4A4 · W4A4 avg of 4 tasks Llama-3.1-8Bavg accuracy: 73.11 -> 74.75Improved
add Hadamard rotation to weight-only RTN (ablation)NVFP4 W4A16 · Llama-3.1-8B W4A16avg accuracy: 77.37 -> 76.41Hurt
add QAT with Hadamard 32 (ablation)MXFP4 Llama-3.2-1B · W4A4 Llama-3.2-1Brecovery %: 68.16 -> 84.55Improved
add QAT to GPTQNVFP4 Llama-3.1-8B · W4A4 real kernelsavg recovery: 95.7 -> 95.4No clear effect
add Hadamard 16 to RTN (ablation)NVFP4 Llama-3.2-3B · W4A4recovery %: 94.45 -> 90.96Hurt
use MXINT4 with MR-GPTQ vs RTNMXINT4 Llama-3.1-8B · W4A4avg accuracy: 63.05 -> 72.51Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 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