AI Paper Atlas

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

CTP:Towards Vision-Language Continual Pretraining via Compatible Momentum Contrast and Topology Preservation

Hongguang Zhu, Yunchao Wei, Xiaodan Liang, Chunjie Zhang, Yao Zhao

ICCV 2023 · Continual learning › Continual fine-tuning of LLMs · approach: Mixture-of-experts for multimodal continual tuning · 35 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 compatible momentum contrast to sequential finetuning (ablation)ALBEF continual pretraining · P9D cross-modal retrievalRm: 56.86 -> 59.86Improved
add cross-modal topology preservation (ablation)ALBEF continual pretraining · P9D cross-modal retrievalRm: 56.86 -> 62.50Improved
add cross-modal topology preservation alone (ablation)ALBEF continual pretraining · P9D multi-modal retrievalmAP@1: 62.15 -> 61.32Hurt
add same-modal topology preservation on top of CMC and Lc (ablation)CTP · P9D cross-modal retrievalRm: 64.38 -> 64.87Improved
use compatible momentum update from both old and current model instead of one side (ablation)CTP · P9D cross-modal retrievalRm: 63.57 -> 64.87Improved
add momentum queue of negatives (ablation)CTP · P9D cross-modal retrievalRm: 63.94 -> 64.87Improved
suppress same-modal diagonal similarity (ablation)CTP · P9D cross-modal retrievalRm: 63.12 -> 64.87Improved
add ER replay buffer (10000 pairs) to CTP (ablation)CTP · P9D cross-modal retrievalRm: 64.87 -> 70.63Improved
replace parameter regularization (EWC) with CTPmemory-free continual VLP · P9D cross-modal retrievalRm: 59.24 -> 64.87Improved
continual methods on multi-modal retrieval versus sequential finetuningCTP and baselines · P9D multi-modal retrievalmAP@1: 62.15 -> 62.64No clear effect
vary momentum coefficient m from 0.7 to 0.99 (ablation)CTP · P9D cross-modal retrievalRm: 64.83 -> 64.74No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 0 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