AI Paper Atlas

Research map / Continual learning / Online and replay-based

Open-World Dynamic Prompt and Continual Visual Representation Learning

Youngeun Kim, Jun Fang, Qin Zhang, Zhaowei Cai, Yantao Shen, Rahul Duggal, Dripta S. Raychaudhuri, Zhuowen Tu, Yifan Xing, Onkar Dabeer

ECCV 2024 · Continual learning › Online and replay-based · approach: Memory-based and Bayesian online learning · 4 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
replace static prompt pool with dynamic prompt generation network (frozen backbone) (ablation)CodaPrompt · 100 CL stages avgRecall@1: 67.95 -> 77.05Improved
add stage tokens to DPG (ablation)DPG · 100 CL stages avgRecall@1: 76.26 -> 77.05Improved
FIFO versus FILO or random replacement of stage tokens (ablation)DPG · 100 CL stages avgRecall@1: 77.05 vs 76.83 vs 77.01No clear effect
cap stage tokens at 5 instead of keeping all 100 (ablation)DPG · 100 CL stages avgRecall@1: 77.14 -> 77.05No clear effect
use low-rank (256) mapping function instead of full-rank (ablation)DPG · In-Shop 10 stagesRecall@1: 71.40 -> 84.09Improved
fine-tune entire backbone instead of using DPG with frozen backbone (ablation)DPaRL · 10 CL stages 4-dataset avgRecall@1: 79.36 -> 69.58Hurt
update backbone with LoRA jointly with DPG (ablation)DPG · 10 CL stages 4-dataset avgRecall@1: 79.36 -> 80.80Improved
use LoRA rather than Adapter for backbone updating in DPaRL (ablation)DPaRL · 10 CL stages 4-dataset avgRecall@1: 78.70 -> 80.80Improved
DPaRL versus prior best prompt-based CL in 10 stagesCodaPrompt · 4 open-world retrieval benchmarks avgRecall@1: 76.11 -> 80.80Improved
DPaRL versus prior prompt methods in closed-world medical domainCoda · MedMNISTtop-1 accuracy: 54.49 -> 60.99Improved

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