AI Paper Atlas

Research map / Architectures and efficiency / Quantization

AutoQRA: Joint Optimization of Mixed-Precision Quantization and Low-rank Adapters for Efficient LLM Fine-Tuning

Changhai Zhou, Shiyang Zhang, Yuhua Zhou, Qian Qiao, Jun Gao, Cheng Jin, KAIZHOU QIN, WEIZHONG ZHANG

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
jointly search bit-width and LoRA rank (AutoQRA <=4-bit) instead of uniform 4-bit QLoRAAutoQRA (<=4-bit) · Task average, LLaMA3.1-8BAvg accuracy (%): 67.45 -> 69.83Improved
allow mixed precision above 4-bit average (AutoQRA Opt) vs FP16 LoRAAutoQRA (Opt) · Task average, Qwen2.5-7BAvg accuracy (%): 71.33 -> 73.19Improved
joint bit-rank search instead of decoupled AMQ then LoRAAutoQRA (<=4-bit) · Task average, LLaMA3.2-3BAvg accuracy (%): 63.31 -> 65.58Improved
joint search instead of AMQ+LoRA on code generationAutoQRA (<=4-bit) · HumanEvalpass@1: 24.40 -> 41.96Improved
use surrogate-guided promotion instead of ranking by low-fidelity scores (ablation)AutoQRA Phase I · Top-3 promotion hit ratemean hit rate: 44.7% -> 67.3%Improved
use AutoQRA search instead of random search (ablation)AutoQRA · Evaluations to reach target accuracynumber of HF evaluations: 107 -> 6Improved
move allocation granularity from layer-level to tensor-level (ablation)AutoQRA · Qwen2.5-3BAvg accuracy (%): 66.33 -> 66.72 (search 55 min -> 3.6 h)No clear effect
allocate per layer instead of uniform model-level (ablation)AutoQRA · Qwen2.5-3BAvg accuracy (%): 62.89 -> 66.33Improved
reduce memory budget to 0.7x (ablation)AutoQRA · Qwen2.5-3B, combined avg with HumanEvalAvg: 64.55 -> 63.92Hurt
reuse configuration searched on Alpaca for HC3 instead of searching on HC3 (ablation)AutoQRA · HC3, Qwen2.5-3BAvg: 65.93 -> 65.54No clear effect
use tuned search hyperparameters (wider promotion set, more Phase II iterations) (ablation)AutoQRA · Qwen2.5-3B 5-task average5-task Avg accuracy (%): 69.07 -> 70.02Improved
remove global Phase I search, importance warm start, multi-fidelity, or optimize only one axis (ablation)AutoQRA · Accuracy vs memory (Figure 6)accuracy-memory frontierHurt

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