AI Paper Atlas

Research map / Architectures and efficiency / Quantization

ParoQuant: Pairwise Rotation Quantization for Efficient Reasoning LLM Inference

Yesheng Liang, Haisheng Chen, Song Han, Zhijian Liu

ICLR 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 9 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 8 independent rotations to channel-wise scaling (ablation)ParoQuant (w/ Stage 2) · C4 perplexity, LLaMA-3-8Bperplexity: 7.41 -> 7.27Improved
add channel-wise scaling to 8 independent rotations (ablation)ParoQuant (w/ Stage 2) · C4 perplexity, LLaMA-3-8Bperplexity: 7.40 -> 7.27Improved
add Stage 2 fine-tuning (ablation)ParoQuant (8 IR + S) · C4 perplexity, LLaMA-3-8Bperplexity: 7.35 -> 7.27Improved
add transforms (8 IR + S) to RTN without stage 2 (ablation)RTN · C4 perplexity, LLaMA-3-8Bperplexity: 7.56 -> 7.35Improved
add Stage 2 when no transform is used (ablation)RTN · C4 perplexity, LLaMA-3-8Bperplexity: 7.56 -> 7.42Improved
increase independent rotations from 0 to 8 (2048 samples) (ablation)ParoQuant · C4 perplexity, LLaMA-3-8Bperplexity: 7.41 -> 7.27Improved
increase independent rotations from 4 to 8 (ablation)ParoQuant · C4 perplexity, LLaMA-3-8Bperplexity: 7.27 -> 7.27No clear effect
increase calibration samples from 128 to 2048 (ablation)ParoQuant · C4 perplexity, LLaMA-3-8Bperplexity: 7.30 -> 7.27Improved
use mixed calibration set instead of RedPajama only (ablation)ParoQuant · MMLU-Pro, Qwen3-4Baccuracy: 69.5 -> 70.1Improved
use ParoQuant instead of AWQReasoning LLMs (4-bit) · Reasoning averageaccuracy: 59.5 -> 61.9Improved
use ParoQuant instead of EfficientQATReasoning LLMs (4-bit) · Reasoning averageaccuracy: 55.6 -> 61.9Improved
use ParoQuant instead of AWQ for decoding throughputQwen3-4B · Decoding throughputtokens/s: 176 -> 160Hurt

Later papers that compared with this method

2 later accepted papers report ParoQuant++ as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.

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