AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

SumRA: Parameter Efficient Fine-tuning with Singular Value Decomposition and Summed Orthogonal Basis

Chin Yuen Kwok, Yongsen Zheng, Jia Qi Yip, Kwok-Yan Lam, Eng Siong Chng

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
use SumRA summed-singular-vector init of A instead of LoRALoRA rank 32, whisper-large-v2 · Common Voice eoWER: 14.42 -> 12.41Improved
use SumRA init instead of LoRALoRA rank 32, whisper-small · Common Voice eoWER: 23.39 -> 20.77Improved
use SumRA instead of PiSSArank 32, whisper-large-v2 · Common Voice iaWER: 8.82 -> 8.17Improved
use SumRA instead of CorDArank 32, whisper-large-v2 · Common Voice iaWER: 9.18 -> 8.17Improved
interleave sum instead of block sum for summing singular vectors (ablation)SumRA, whisper-small rank 32 · Common Voice eoWER: 21.68 -> 20.77Improved
interleave sum instead of block sum for summing singular vectors (ablation)SumRA, whisper-small rank 32 · Common Voice kmrWER: 47.30 -> 44.47Improved
greedy sum instead of interleave sum (ablation)SumRA, whisper-small rank 32 · Common Voice fy-NL and mhrWER: 33.37 -> 33.91 (fy-NL), 36.30 -> 37.53 (mhr), 13.38 -> 13.16 (ia)Mixed
train matrix A as well as B (ablation)SumRA, whisper-small · Esperanto 10hWER: 20.77 -> 20.14Improved
train matrix A as well as B (ablation)SumRA, whisper-small · Esperanto 50hWER: 14.49 -> 13.75Improved
train matrix A as well as B (ablation)SumRA, whisper-small · Esperanto 100hWER: 13.39 -> 13.02Improved
SumRA instead of LoRA with more datawhisper-small · Esperanto 100hWER: 13.28 -> 13.39No clear effect
SumRA instead of LoRA with 50h datawhisper-small · Esperanto 50hWER: 15.20 -> 14.49Improved

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