AI Paper Atlas

Research map / Continual learning / Online and replay-based

Decomposing Prompts, Composing Actions: A Multi-Granularity Prompting Approach for Incremental Action Learning

Xinyi Cheng, Chenghao Xu, Xi Wang, Jiexi Yan, Yanhua Yang

AAAI 2026 · Continual learning › Online and replay-based · approach: Dynamic expansion and isolation · 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
add temporal prompt pool Pt (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 46.8 -> 57.7Improved
add spatial prompt pool Ps (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 52.0 -> 57.7Improved
use multi-granularity prompts instead of single comprehensive prompt (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 40.6 -> 57.7Improved
add task-agnostic prompt pool Pa (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 54.1 -> 57.7Improved
add Dissimilarity Attention Rectification (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 56.7 -> 57.7Improved
add 3 prompts per task instead of 2 (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 57.0 -> 57.7Improved
add 5 prompts per task instead of 3 (ablation)DPCA · NTU RGB+D CS, final sessionAHM: 57.7 -> 56.2Hurt
use DPCA instead of POET on PKU-MMDPOET · PKU-MMDAHM: 55.2 -> 56.5Improved
use DPCA instead of POET on SHREC-2017POET · SHREC-2017AHM: 55.8 -> 57.3Improved
use DPCA instead of POET with ST-GCN backbonePOET · NTU RGB+D CS (ST-GCN)AHM: 54.3 -> 55.7Improved
use DPCA instead of POET for new-class learningPOET · NTU RGB+D CS, final sessionNew-class accuracy: 55.8 -> 55.3No clear effect
use image-style prompt methods (L2P, CODA-P) for skeleton continual learningfrozen CTR-GCN · NTU RGB+D CS, final sessionAHM: 57.7 -> 0.0 (L2P)Hurt

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