AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

QWHA: Quantization-Aware Walsh-Hadamard Adaptation for Parameter-Efficient Fine-Tuning on Large Language Models

Hyesung Jeon, Seojune Lee, Beomseok Kang, Yulhwa Kim, Jae-Joon Kim

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Efficient LLM fine-tuning systems · 1 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 AdaAlloc parameter selection (vs random with QA init) (ablation)QWHA (WHA) · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 24.48 -> 29.11Improved
add AdaAlloc parameter selection (vs magnitude) (ablation)QWHA (WHA) · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 28.12 -> 29.11Improved
add AdaAlloc parameter selection (vs LoCA selection) (ablation)QWHA (WHA) · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 21.15 -> 29.11Improved
add AdaAlloc parameter selection (vs SSH selection) (ablation)QWHA (WHA) · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 27.14 -> 29.11Improved
add quantization-aware initialization and refinement to WHA with random selection (ablation)WHA · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 24.41 -> 24.48No clear effect
use WHA adapter instead of DCA (ablation)QWHA with AdaAlloc · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 27.29 -> 29.11Improved
use WHA adapter instead of DHA (ablation)QWHA with AdaAlloc · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 27.52 -> 29.11Improved
use WHA adapter instead of plain sparse adapter (ablation)QWHA with AdaAlloc · GSM8k 2-bit LLaMA-3.2-3Baccuracy (%): 26.54 -> 29.11Improved
add QWHA at 2-bit instead of CLoQ low-rank adapterLLaMA-3.1-8B · GSM8k 2-bitaccuracy (%): 33.89 -> 37.83Improved
add QWHA at 2-bit instead of CLoQ low-rank adapterLLaMA-3.1-8B · CSQA 2-bitaccuracy (%): 56.49 -> 60.98Improved
add quantization-aware init (QWHA vs SSH without it)LLaMA-3.1-8B · GSM8k 3-bitaccuracy (%): 50.34 -> 55.34Improved
use QWHA at 4-bit vs 16-bit fine-tuned modelLLaMA-3.1-8B · GSM8k 4-bitaccuracy (%): 59.74 -> 56.10Hurt

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