Research map / Architectures and efficiency / Quantization
CodeQuant: Unified Clustering and Quantization for Enhanced Outlier Smoothing in Low-Precision Mixture-of-Experts
Xiangyang Yin, Xingyu Liu, Tianhua Xia, BO BAO, Vithursan Thangarasa, Valavan Manohararajah, Eric Sather, Sai Qian Zhang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace random rotation with AOS fine-tuned rotation (ablation) | CodeQuant · DeepSeek-V2-Lite A4W4 Embedding-wise | average accuracy: 0.652 -> 0.667; perplexity down 0.23 on WikiText2 and 0.31 on C4 | Improved |
| add KL divergence penalty to centroid fine-tuning (ablation) | CodeQuant ACCF · DeepSeek-V2-Lite A4W4 Block-wise | average accuracy: 0.658 -> 0.667 | Improved |
| add KL divergence penalty to centroid fine-tuning (ablation) | CodeQuant ACCF · Phi-mini-MoE-Instruct A4W4 Block-wise | average accuracy: 0.694 -> 0.700 | Improved |
| add KL penalty to reduce router change (ablation) | CodeQuant · DeepSeek-V2-Lite router top-K change rate | change rate (%): 24.33 -> 22.82 | Improved |
| add KL penalty to reduce router change (ablation) | CodeQuant · Qwen3-30B-A3B router top-K change rate | change rate (%): 60.21 -> 59.58 | Improved |
| add permutation (POG) before block-wise clustering (ablation) | CodeQuant · DeepSeek-V2-Lite A4W4 Block-wise g=1024 | average accuracy: 0.663 -> 0.668 | Improved |
| add permutation (POG) before block-wise clustering (ablation) | CodeQuant · Phi-mini-MoE-Instruct A4W4 Block-wise g=1024 | average accuracy: 0.710 -> 0.714 | Improved |
| use CodeQuant clustering instead of SqueezeLLM at 2-bit weights (ablation) | CodeQuant · DeepSeek-V2-Lite A4W2 | average accuracy: 0.496 -> 0.568 | Improved |
| reduce weight bits from 4 to 2 (A4W4 to A4W2) (ablation) | CodeQuant · DeepSeek-V2-Lite | average accuracy: 0.667 -> 0.568 | Hurt |
| use CodeQuant instead of QuaRot | quantized Qwen3-30B-A3B · A4W4 Embedding-wise | average accuracy: 0.581 -> 0.694 | Improved |
| use CodeQuant instead of QuaRot | quantized Mixtral-8x7B · A4W4 Embedding-wise | WikiText2 perplexity: 16.79 -> 4.65 | Improved |
| use CodeQuant instead of QuaRot | quantized Qwen3-30B-A3B · GSM8K 8-shot A4W4 | accuracy: 0.508 -> 0.867 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- QuaRot — QuaRot: Outlier-Free 4-Bit Inference in Rotated LLMs (NeurIPS 2024) · better on 10 of 10 shared results
- DuQuant — DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs (NeurIPS 2024) · better on 2 of 2 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
- DuQuant: Distributing Outliers via Dual Transformation Makes Stronger Quantized LLMs (NeurIPS 2024 · similarity 0.89)
- MoEQuant: Enhancing Quantization for Mixture-of-Experts Large Language Models via Expert-Balanced Sampling and Affinity Guidance (ICML 2025 · similarity 0.89)
- Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models (ICLR 2024 · similarity 0.89)
- LeanQuant: Accurate and Scalable Large Language Model Quantization with Loss-error-aware Grid (ICLR 2025 · similarity 0.88)
- ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals (ICML 2025 · similarity 0.87)
- MxMoE: Mixed-precision Quantization for MoE with Accuracy and Performance Co-Design (ICML 2025 · similarity 0.87)
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.
