AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Learning in the Fisher Subspace: A Guided Initialization for LoRA Fine-Tuning

Zhi-Quan Feng, Ying-Jia Lin, Hung-Yu Kao

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: LoRA variants and theory · 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
use low-Fisher-energy directions with Fisher-based scaling instead of SVD singular-value scaling (min directions) (ablation)FILet · Reasoning avg, Llama2-7Baccuracy (%): 81.9 -> 83.1Improved
use Fisher-based scaling instead of SVD scaling (ablation)FILet · Reasoning avg, Llama3-8Baccuracy (%): 85.8 -> 86.7Improved
select minimal-Fisher-energy directions instead of maximal (ablation)FILet · Reasoning avg, Llama2-7Baccuracy (%): 81.8 -> 83.1Improved
select minimal-Fisher-energy directions instead of random (ablation)FILet · Reasoning avg, Llama3-8Baccuracy (%): 86.4 -> 86.7Improved
select minimal-Fisher-energy directions instead of random (ablation)FILet · Image classification avg, ViT-B/32accuracy (%): 68.9 -> 69.3Improved
use minimal (rather than maximal) singular-value scaling in preliminary directions test (ablation)LoRA init on Llama2-7B · ARC-c and BoolQaccuracyImproved
increase initialization minibatch size from 80 to Full dataset (ablation)FILet · Reasoning avgaccuracy (%): up to 0.7% on Llama2-7B and 0.9% on Llama3-8BImproved
increase LoRA rank from 32 to 128 (ablation)FILet · Reasoning avg, Llama2-7Baccuracy (%): 83.1 -> 82.6Hurt
increase LoRA rank from 4 to 64 (ablation)FILet · Reasoning avg, Llama2-7Baccuracy (%): 80.8 -> 83.2Improved
replace SVD-based or data-driven init (KaSA, LoRA-Dash) with Fisher-guided initLoRA · GSM8k Llama3-8Baccuracy (%): 55.8 -> 58.8Improved
replace baselines with Fisher-guided init on ARC-cLoRA · ARC-c Llama3-8Baccuracy (%): 80.3 -> 79.9Hurt
Fisher-guided init on image classification vs LoRALoRA · ViT-B/32 avgaccuracy (%): 68.4 -> 69.3Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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