AI Paper Atlas

Research map / Architectures and efficiency / Quantization

NanoQuant: Efficient Sub-1-Bit Quantization of Large Language Models

Hyochan Chong, Dongkyu Kim, Changdong Kim, Minseop Choi

ICML 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 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 error mitigation to block reconstruction (ablation)NanoQuant on Qwen3-8B-Base · WikiText-2perplexity: 206.03 -> 15.07Improved
add factorized component refinement to initialization (ablation)NanoQuant on Qwen3-8B-Base · WikiText-2perplexity: 206.03 -> 15.00Improved
add factorized refinement on top of error mitigation (ablation)NanoQuant on Qwen3-8B-Base · WikiText-2perplexity: 15.07 -> 13.58Improved
add factorized refinement on top of error mitigation (ablation)NanoQuant on Qwen3-8B-Base · zero-shot accuracyaverage accuracy (%): 46.40 -> 46.75No clear effect
add scale-only model reconstruction (ablation)NanoQuant on Qwen3-8B-Base · WikiText-2perplexity: 13.58 -> 12.47Improved
add scale-only model reconstruction (ablation)NanoQuant on Qwen3-8B-Base · zero-shot accuracyaverage accuracy (%): 46.75 -> 48.94Improved
use LB-ADMM initialization instead of Dual-SVID (ablation)NanoQuant on Rnj-1 at 0.8 bits · WikiText-2perplexity: 167.73 -> 20.06Improved
use LB-ADMM initialization instead of DBF ADMM (ablation)NanoQuant on Rnj-1 at 0.8 bits · WikiText-2perplexity: 30.27 -> 20.06Improved
use LB-ADMM initialization instead of DBF ADMM (ablation)NanoQuant on Rnj-1 at 0.8 bits · zero-shot accuracyaverage accuracy (%): 37.20 -> 39.29Improved
use 2.10M calibration tokens instead of 0.26MNanoQuant on Llama-2-7B 1-bit · WikiText-2perplexity: 10.34 -> 8.85Improved
compress Llama-2-7B to 0.55 bit instead of 1 bitNanoQuant · WikiText-2perplexity: 10.34 -> 16.66Hurt
use NanoQuant PTQ instead of QAT LittleBitQwen3-4B-Base 1-bit · WikiText-2perplexity: 14.79 -> 12.62Improved

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