AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Multi-Task Low-Rank Model Adaptation

Zichen Tian, Antoine Ledent, Qianru Sun

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: LoRA variants and theory

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 block-level adaptation (ablation)HydraLoRA · average of DOTA, iNat2018, Dolly-15k, BBHaverage accuracy (%): 61.1 -> 63.0Improved
apply block-level adapters to attention and FFN instead of Wq, Wv (ablation)HydraLoRA · average of DOTA, iNat2018, Dolly-15k, BBHaverage accuracy (%): 61.1 -> 63.1Improved
add spectral-aware regularization (ablation)mtLoRA block-level · average of DOTA, iNat2018, Dolly-15k, BBHaverage accuracy (%): 63.2 -> 63.8Improved
add fine-grained routing (ablation)mtLoRA block-level · average of DOTA, iNat2018, Dolly-15k, BBHaverage accuracy (%): 63.2 -> 63.1Mixed
add all three designs (ablation)HydraLoRA · average of DOTA, iNat2018, Dolly-15k, BBHaverage accuracy (%): 61.1 -> 63.9Improved
increase routing groups from g=1 to g=32 (ablation)HydraLoRA · BBHaccuracy (%): 35.5 -> 37.7Improved
increase routing groups from g=1 to g=2 (ablation)HydraLoRA · BBHaccuracy (%): 35.5 -> 37.0Improved
add orthogonal regularization without routing (ablation)HydraLoRA uniform routing · DOTAaccuracy (%): 20.5Hurt
add orthogonal regularization with dynamic routing (ablation)HydraLoRA · DOTAaccuracy (%): 89.0 -> 89.8Improved
add sparsity regularization with dynamic routing (ablation)HydraLoRA · DOTAaccuracy (%): 89.0 -> 87.9Hurt
raise orthogonal regularization strength from 0.25 to 1.0 (ablation)HydraLoRA with routing · Flan-v2 to BBHaccuracy (%): -1.8%Hurt
use block-level instead of component-level adapters (ablation)LoRA · gradient cosine similarity between taskscosine similarity: -0.054 -> -0.013Improved

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