AI Paper Atlas

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

PACE: Pretrained Audio Continual Learning

Chang Li, Kanglei Zhou, Liyuan Wang

ICLR 2026 · Continual learning › Continual fine-tuning of LLMs · approach: Continual pretraining of language models · 0 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 improved first-session adaptation (FSA) (ablation)PACE · TIMIT-2 / TIMIT-3 / VocalSetaccuracy: 75.87 -> 90.95; 75.87 -> 94.05; 61.51 -> 69.08Improved
add multi-session adaptation (MSA) (ablation)PACE · TIMIT-2 / TIMIT-3 / VocalSetaccuracy: 85.63 -> 90.95; 89.92 -> 94.05; 62.82 -> 69.08Improved
add boundary-aware regularization L_reg (ablation)PACE · TIMIT-2 / TIMIT-3 / VocalSetaccuracy: 89.21 -> 90.95; 93.73 -> 94.05; 66.78 -> 69.08Improved
add gradient projection (GP) in MSA (ablation)PACE · TIMIT-2 / TIMIT-3 / VocalSetaccuracy: 88.01 -> 90.95; 89.05 -> 94.05; 58.55 -> 69.08Improved
use improved FSA instead of naive FSA (ablation)PACE FSA · SC2accuracy: 90.53 -> 91.87Improved
add low learning rate for FSA head (ablation)FSA · ESC-50accuracy: 92.25 -> 93.75Improved
add early head freezing to low learning rate (ablation)FSA · ESC-50accuracy: 93.75 -> 94.50Improved
add layer-aware tuning of later layers (ablation)FSA · ESC-50accuracy: 94.50 -> 95.75Improved
add naive FSA to RanPAC (ablation)RanPAC · ESC-50accuracy: 92.50 -> 92.25Hurt
add FSA to RanPAC (ablation)RanPAC · SC2accuracy: 81.22 -> 90.53Improved
add gradient projection to MSA (ablation)MSA · VocalSetforgetting: 57.90 -> 23.69Improved
replace prompt-based CL with PACEHiDe-Prompt · TIMIT-2accuracy: 47.78 -> 90.95Improved

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