AI Paper Atlas

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

Mind the Interference: Retaining Pre-trained Knowledge in Parameter Efficient Continual Learning of Vision-Language Models

Longxiang Tang, Zhuotao Tian, Kai Li, Chunming He, Hantao Zhou, Hengshuang Zhao, Xiu Li, Jiaya Jia

ECCV 2024 · Continual learning › Continual fine-tuning of LLMs · approach: Low-rank and model-merging adaptation · 26 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 zero-initialization to prompt tuning (without residual attention) (ablation)prompt tuning · MTILTransfer: 57.7 -> 57.3No clear effect
replace prompt prepending with residual attention (ablation)DIKI · MTILTransfer: 57.7 -> 59.9Improved
add zero-initialization to residual attention (ablation)DIKI · MTILTransfer: 59.9 -> 63.1Improved
add distribution-aware calibration (ablation)DIKI · MTILTransfer: 63.1 -> 68.7Improved
add zero-init and calibration (ablation)DIKI · MTILLast: 85.2 -> 85.1No clear effect
increase uniform initialization bound of residual parametersDIKI · MTILTransferHurt
DIKI vs ZSCL with 0.86% trainable params and no extra dataCLIP ViT-B/16 · MTILLast: 83.6 -> 85.1Improved
DIKI vs prompt methods on zero-shot retentionCLIP ViT-B/16 · MTILTransfer: 52.2 -> 68.7Improved
replace prompt prepending with IKI in DualPromptDualPrompt · 10-split CIFAR-100Avg. Acc: 86.51 -> 88.77Improved
replace prompt prepending with IKI in L2PL2P · 10-split CIFAR-100Forgetting: 7.35 -> 7.28No clear effect
use fixed larger calibration weight instead of sample-adaptive weightDIKI · MTIL first task Aircraftzero-shot ability vs trained-task accuracyMixed

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