AI Paper Atlas

Research map / Architectures and efficiency / Quantization

ReQAT: Achieving Full-Precision Reasoning Accuracy with 4-bit Floating-Point Quantization-Aware Training

Janghwan Lee, Sihwa Lee, Jinseok Kim, Yongjik Kim, Jieun Lim, Jinwook Oh, Jungwook Choi

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 trace alignment to two-stage QAT (ablation)FT+QAT · MXFP4 W4A16 R1-Qwen-14B (350M)AIME accuracy: 62.60 -> 67.29Improved
add Q-FIT (pre-RoPE scaling + post-RoPE shifting) to TAQ (ablation)ReQAT_T · NVFP4 W4A4KV4 R1-Qwen-14B (280M)AIME accuracy: 63.13 -> 65.94Improved
use only one of scale or shift in Q-FIT (ablation)ReQAT_TQ · NVFP4 W4A4KV4 R1-Qwen-14BAIME accuracy: 65.94 -> 62.71Hurt
add SEM to TAQ+Q-FIT (ablation)ReQAT_TQ · R1-Llama-8B AIME-120 NVFP4 W4A4KV4AIME accuracy: 40.32 -> 41.85Improved
add SEM to TAQ+Q-FIT (ablation)ReQAT_TQ · R1-Llama-8B GSM8K NVFP4 W4A4KV4accuracy: 89.86 -> 89.85No clear effect
add SEM to TAQ+Q-FIT (ablation)ReQAT_TQ · R1-Llama-8B MATH-500 NVFP4 W4A4KV4accuracy: 90.72 -> 90.53No clear effect
use soft weighting instead of binary mask in SEM (ablation)SEM · MXFP4 W4A4 AIME-90 (210M)AIME accuracy: 63.19 -> 65.14Improved
train more epochs of TAQ instead of adding SEM (ablation)multi-epoch TAQ · R1-Llama-8B AIME NVFP4AIME accuracy: 38.34 -> 37.71No clear effect
add TAQ over standard QATReQAT_T vs QAT · NVFP4 W4A4KV4 R1-Qwen-14B (280M)AIME accuracy: 58.86 -> 63.13Improved
add more BF16 FT epochs before PTQ (ablation)FT then PTQ · R1-Llama-8B NVFP4 W4A4KV4AIME accuracy: 11.05 -> 15.42Hurt
add QAT after fine-tuning without trace alignment (ablation)FT+QAT · MXFP4 W4A16 R1-Qwen-14B (350M)AIME accuracy: 62.29 -> 62.60No clear effect
quantize with ReQAT vs BF16 FTReQAT · NVFP4 W4A4KV4 R1-Qwen-14BAIME accuracy: 65.46 -> 65.94Improved

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