AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

On-the-Fly Adaptation to Quantization: Configuration-Aware LoRA for Efficient Fine-Tuning of Quantized LLMs

Rongguang Ye, Ming Tang, Edith C. H. Ngai

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging

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
use a configuration-aware adjustment model instead of one shared LoRA adapterShared-LoRA · SST-2 (RoBERTa-Large)avg accuracy gap vs QLoRA: -5.06% -> +1.74%Improved
use CoA-LoRA instead of per-configuration QLoRA fine-tuningQLoRA · QNLI/MNLI/SST-2/QQPtotal training time: QNLI 119m -> 57mImproved
use CoA-LoRA instead of QLoRAQLoRA · MNLI (RoBERTa-Large)avg accuracy gap: +8.89%Improved
add Pareto-based configuration search to training configuration set (ablation)CoA-LoRA · QNLI and SST-2accuracy: nearly 2%Improved
randomly add ten new configurations per epoch instead of guided search (ablation)CoA-LoRA · configuration search ablationaccuracyNo clear effect
use segmented Pareto selection (U=20 or 40) instead of none (U=0) (ablation)CoA-LoRA · QNLI/SST-2/QQPaccuracyImproved
increase segments from U=20 to U=40 (ablation)CoA-LoRA · SST-2 vs QNLI/QQPaccuracyMixed
increase LoRA rank (32, 64, 128) (ablation)CoA-LoRA · four GLUE tasksHV: QNLI 0.619 / 0.629 / 0.623No clear effect
use CoA-LoRA on integer mixed-precision quantizationQLoRA · QNLI/MNLI/SST-2/QQPaccuracyMixed
use CoA-LoRA on LLaMA-2-13B at 3 to 4 bitsQLoRA · C4 perplexity, rank 64perplexity: 4-bit: 6.89 -> 6.99No clear effect
use CoA-LoRA at 2.5 bits on LLaMA-2-13BQLoRA · C4 perplexity, 2.5-bitperplexity: 9.11 -> 8.65Improved
use CoA-LoRA at zero-shot on Qwen2.5-3BLQ-LoRA · zero-shot 3-bitaverage accuracy: 51.90 -> 53.35Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most 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

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