AI Paper Atlas

Research map / Continual learning / Continual fine-tuning of LLMs

Spectral Imbalance Causes Forgetting in Low-Rank Continual Adaptation

Hao Gu, Mao-Lin Luo, Zi-Hao Zhou, Han-Chen Zhang, Min-Ling Zhang, Tong Wei

ICML 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Low-rank and model-merging adaptation

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 EBLoRA (balanced energy + orthogonality)LoRA-FT, LLaVA-1.5-7B · UCITMFN: 61.4 -> 72.8Improved
replace LoRA with EBLoRALoRA-FT, LLaVA-1.5-7B · MLLM-DCLMFN: 54.5 -> 66.7Improved
add energy-balanced optimization (s*U*V factorization) to LoRA (ablation)LoRA-FT · UCITBWT: -15.4 -> -5.1Improved
add energy-balanced optimization to LoRA (ablation)LoRA-FT · UCITFWT: 26.8 -> 34.3Improved
add gradient orthogonality to EBO (ablation)EBO · UCITBWT: -5.1 -> -2.6Improved
add gradient orthogonality to EBO (ablation)EBO · UCITFWT: 34.3 -> 34.0No clear effect
add depth-aware initialization of s (ablation)EBO+GO · UCITMFN: 72.1 -> 72.8Improved
replace scalar s with learnable diagonal S (free spectrum) (ablation)EBO · UCITMFN: 70.2 -> 68.1Hurt
smooth singular values of LoRA adapters before merging (ablation)LoRA merging · UCITNAIImproved
use Newton-Schulz orthogonalization instead of whitening retraction (ablation)EBLoRA · UCITMFN: 72.8 -> 71.6Hurt
use EBLoRA instead of LoRAtraining cost · UCITtraining time (s/it): 3.28 -> 4.20Hurt

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