AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Efficiently Seeking Flat Minima for Better Generalization in Fine-Tuning Large Language Models and Beyond

Jiaxin Deng, Qingcheng Zhu, Junbiao Pang, Linlin Yang, Zhongqian Fu, Baochang Zhang

AAAI 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: LoRA variants and theory · 2 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
add transferred full-space SAM perturbation (FMLoRA) to LoRALoRA · RoBERTa-large few-shot avgaverage score: 80.0 -> 83.1Improved
add EMA-estimated perturbation (EFMLoRA) to LoRALoRA · RoBERTa-large few-shot avgaverage score: 80.0 -> 82.3Improved
use EFMLoRA instead of FMLoRA with exact second pass (ablation)FMLoRA · RoBERTa-large few-shot avgaverage score: 83.1 -> 82.3Hurt
add EFMLoRA to LoRALoRA · GLUE (8 tasks avg, RoBERTa-large)average score: 88.4 -> 89.4Improved
use EFMLoRA instead of full fine-tuningfull fine-tuning · GLUE (RoBERTa-large avg)average score: 88.9 -> 89.4Improved
add EFMLoRA to LoRALoRA · E2E NLG (GPT-2 Medium)BLEU: 69.2 -> 69.7Improved
add FMLoRA to LoRALoRA · E2E NLG (GPT-2 Medium)BLEU: 69.2 -> 69.2No clear effect
add EFMLoRA to LoRALoRA · E2E NLG (GPT-2 Large)CIDEr: 2.53 -> 2.52No clear effect
add EFMLoRA to LoRALoRA · CLIP EuroSAT 1-shotaccuracy (%): 72.3 -> 78.3Improved
add EFMLoRA to LoRALoRA · CLIP Flowers102 1-shotaccuracy (%): 83.2 -> 81.0Hurt
add EFMLoRA to LoRALoRA · CLIP EuroSAT 16-shotaccuracy (%): 92.1 -> 91.6No clear effect
increase perturbation radius rho from 0.2 to 0.6 (ablation)EFMLoRA · ScienceQA (Qwen-VL-Chat)accuracy (%): 91.6 -> 89.6Hurt

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