AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

TuckA: Hierarchical Compact Tensor Experts for Efficient Fine-Tuning

Qifeng Lei, Zhiyong Yang, Qianqian Xu, Cong Hua, Peisong Wen, Qingming Huang

AAAI 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning · 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
replace LoRA with TuckA (multi-expert Tucker adapter, r=8,k=2)DeBERTa-v3-base · GLUE averageaverage: 86.63 -> 88.73Improved
use TuckA r=2,k=4 (0.16M) vs LoRA (1.19M)DeBERTa-v3-base · GLUE averageaverage: 86.63 -> 88.56Improved
replace HRA with TuckA r=8,k=2DeBERTa-v3-base · GLUE MRPCaccuracy/F1: 92.87 -> 92.82Hurt
replace HRA with TuckA r=6,k=1ViT-Base 10-shot · CIFAR100accuracy: 83.45 -> 84.81Improved
use TuckA instead of HRA/DoRA for GPU memoryViT-Base CIFAR100 · peak GPU memoryMB: HRA 9530 -> TuckA 4514 (LoRA 4419)Improved
replace LoRA with TuckA r=32,k=4Llama2-7B · MATHaccuracy: 6.62 -> 9.02Improved
replace LoRA r=32 with TuckA r=16,k=4 (fewer params)Llama2-7B · GSM-8Kaccuracy: 50.34 -> 50.34No clear effect
increase number of expert groups (ablation)TuckA · CIFAR-100 few-shotaccuracyImproved
increase rank to very large values (ablation)TuckA · CIFAR-100 few-shotaccuracyMixed
unrestrict HRA to single-group TuckA (ablation)TuckA vs HRA · CIFAR100 few-shotaccuracyImproved
use data-aware initialization instead of Kaiming for router centroids (ablation)TuckA · CIFAR-100expert load balanceImproved

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