AI Paper Atlas

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

Mind the Gap: Preserving and Compensating for the Modality Gap in CLIP-Based Continual Learning

Linlan Huang, Xusheng Cao, Haori Lu, Yifan Meng, Fei Yang, Xialei Liu

ICCV 2025 · Continual learning › Continual fine-tuning of LLMs · approach: Low-rank and model-merging adaptation · 22 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 adaptive modality gap preservation (MGP) to simple fine-tuning (ablation)CLIP ViT-B/16 LoRA · ImageNet100 class-incrementalLast accuracy: 72.74 -> 76.86Improved
add modality gap compensation (MGC) to simple fine-tuning (ablation)CLIP ViT-B/16 LoRA · ImageNet100 class-incrementalLast accuracy: 72.74 -> 74.58Improved
add MGC on top of MGP (ablation)CLIP ViT-B/16 LoRA · ImageNet100 class-incrementalLast accuracy: 76.86 -> 78.38Improved
MGP instead of distillation to limit modality gap change (ablation)CLIP ViT-B/16 · ImageNet100 class-incrementalLast accuracy: 75.85 -> 78.38Improved
fine-tune naively without gap control (negative similarity falls) (ablation)CLIP ViT-B/16 · ImageNet100 class-incrementalmean negative cosine similarity: 0.1719 -> 0.0579Hurt
add 2000 replay samples to MG-CLIPCLIP ViT-B/16 · CIFAR-100Last accuracy: 80.40 -> 82.58Improved
add 2000 replay samples to MG-CLIPCLIP ViT-B/16 · ImageNet100 class-incrementalLast accuracy: 78.38 -> 80.74Improved
vary threshold alpha 5% to 30% (ablation)CLIP ViT-B/16 · ImageNet100 class-incrementalLast accuracy: 77.88, 78.38, 77.34, 76.9 for 5%, 10%, 20%, 30%Mixed
vary ensemble weight beta 1 to 8 (ablation)CLIP ViT-B/16 · ImageNet100 class-incrementalLast accuracy: 77.58, 77.92, 78.38, 77.98, 77.32 for 1, 2, 4, 6, 8Mixed
vary LoRA rank 4 to 32 (ablation)CLIP ViT-B/16 · ImageNet100 class-incrementalLast accuracy: 78.02, 78.38, 78.32, 78.26 for 4, 8, 16, 32No clear effect
add EMA (ablation)CLIP ViT-B/16 · ImageNet-R 10-taskLast accuracy: 82.67 -> 82.92No clear effect
add per-task epoch estimation (ablation)CLIP ViT-B/16 · ImageNet-R 10-taskLast accuracy: 82.67 -> 82.82No clear effect

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