Research map / Architectures and efficiency / Quantization
NanoQuant: Efficient Sub-1-Bit Quantization of Large Language Models
Hyochan Chong, Dongkyu Kim, Changdong Kim, Minseop Choi
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 |
|---|---|---|---|
| add error mitigation to block reconstruction (ablation) | NanoQuant on Qwen3-8B-Base · WikiText-2 | perplexity: 206.03 -> 15.07 | Improved |
| add factorized component refinement to initialization (ablation) | NanoQuant on Qwen3-8B-Base · WikiText-2 | perplexity: 206.03 -> 15.00 | Improved |
| add factorized refinement on top of error mitigation (ablation) | NanoQuant on Qwen3-8B-Base · WikiText-2 | perplexity: 15.07 -> 13.58 | Improved |
| add factorized refinement on top of error mitigation (ablation) | NanoQuant on Qwen3-8B-Base · zero-shot accuracy | average accuracy (%): 46.40 -> 46.75 | No clear effect |
| add scale-only model reconstruction (ablation) | NanoQuant on Qwen3-8B-Base · WikiText-2 | perplexity: 13.58 -> 12.47 | Improved |
| add scale-only model reconstruction (ablation) | NanoQuant on Qwen3-8B-Base · zero-shot accuracy | average accuracy (%): 46.75 -> 48.94 | Improved |
| use LB-ADMM initialization instead of Dual-SVID (ablation) | NanoQuant on Rnj-1 at 0.8 bits · WikiText-2 | perplexity: 167.73 -> 20.06 | Improved |
| use LB-ADMM initialization instead of DBF ADMM (ablation) | NanoQuant on Rnj-1 at 0.8 bits · WikiText-2 | perplexity: 30.27 -> 20.06 | Improved |
| use LB-ADMM initialization instead of DBF ADMM (ablation) | NanoQuant on Rnj-1 at 0.8 bits · zero-shot accuracy | average accuracy (%): 37.20 -> 39.29 | Improved |
| use 2.10M calibration tokens instead of 0.26M | NanoQuant on Llama-2-7B 1-bit · WikiText-2 | perplexity: 10.34 -> 8.85 | Improved |
| compress Llama-2-7B to 0.55 bit instead of 1 bit | NanoQuant · WikiText-2 | perplexity: 10.34 -> 16.66 | Hurt |
| use NanoQuant PTQ instead of QAT LittleBit | Qwen3-4B-Base 1-bit · WikiText-2 | perplexity: 14.79 -> 12.62 | Improved |
Closest papers by meaning
- LittleBit: Ultra Low-Bit Quantization via Latent Factorization (NeurIPS 2025 · similarity 0.92)
- ABQ-LLM: Arbitrary-Bit Quantized Inference Acceleration for Large Language Models (AAAI 2025 · similarity 0.91)
- AffineQuant: Affine Transformation Quantization for Large Language Models (ICLR 2024 · similarity 0.91)
- LeanQuant: Accurate and Scalable Large Language Model Quantization with Loss-error-aware Grid (ICLR 2025 · similarity 0.91)
- SpQR: A Sparse-Quantized Representation for Near-Lossless LLM Weight Compression (ICLR 2024 · similarity 0.90)
- BiLLM: Pushing the Limit of Post-Training Quantization for LLMs (ICML 2024 · similarity 0.90)
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.
