AI Paper Atlas

Research map / Continual learning / Online and replay-based

VQACL: A Novel Visual Question Answering Continual Learning Setting

Xi Zhang, Feifei Zhang, Changsheng Xu

CVPR 2023 · Continual learning › Online and replay-based · approach: Unified and unsupervised continual frameworks · 36 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 sample-specific (SS) and sample-invariant (SI) representation learning with rehearsaltransformer encoder-decoder VQA continual learner · VQA v2 standard testingAP (%): +1.78 to +4.74 over rehearsal baselines (ER, DER, VS)Improved
add SS+SI representation learning with rehearsaltransformer encoder-decoder VQA continual learner · NExT-QA standard testingAP (WUPS): +1.72 to +6.1 over rehearsal baselinesImproved
add SS+SI representation learning with rehearsaltransformer encoder-decoder VQA continual learner · VQA v2 and NExT-QA novel composition testingAP (%): +1.62 to +22.57 (VQA v2); +2.02 to +22.31 (NExT-QA) over baselinesImproved
remove sample-specific feature (ablation)full model · VQA v2 Group-1 novel compositionAP (%): 35.50 -> 15.49Hurt
add sample-invariant (SI) feature (ablation)full model · VQA v2 Group-1 novel compositionAP (%): 31.86 -> 35.50Improved
use separate question and visual SI features instead of single fused SI feature (ablation)full model · VQA v2 Group-1 novel compositionAP (%): 32.35 -> 35.50Improved
use separate question and visual SI features instead of single fused SI feature (ablation)full model · VQA v2 standard testingAP (%): 31.88 -> 32.27Improved
add sample-invariant (SI) feature (ablation)full model · VQA v2 standard testingAP (%): 30.55 -> 32.27Improved
switch regularization (EWC, MAS) to rehearsal methodstransformer encoder-decoder baseline · VQA v2 standard testingAP (%): EWC 15.77 / MAS 20.56 vs ER 36.99Improved
add DER and VS rehearsal instead of ERtransformer encoder-decoder baseline · NExT-QA standard testingAP (WUPS): ER 30.55 vs DER 26.17, VS 28.13Hurt
increase memory size from 1000 to 5000 (VQA v2) or 100 to 500 (NExT-QA)rehearsal methods · VQA v2 and NExT-QA standard and composition testingAP (%)Mixed
change linguistic task order (CW, DL, TN first)Vanilla model · NExT-QAAP (%): 0.42 to 2.19 variation in last-task APNo clear effect

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