Research map / Continual learning / Continual fine-tuning of LLMs
PACT: Phase-Like Transition Constraints in Adapter-Based Continual Learning of Vision-Language Models
Xuan Wang, Guiguang Ding, Jungong Han
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace per-task LoRA with MLP-structured adapters using only SAI (1)(2) (no gate) (ablation) | PACT · MTIL Order-I | Last accuracy: 79.54 -> 9.83 | Hurt |
| add convex gating (SAI 4) to the adapter (ablation) | PACT · MTIL Order-I | Last accuracy: 9.83 -> 85.60 | Improved |
| add all components (SAI cloning, Lmarg, Lstab, post-convergence) on top of gated adapter (ablation) | PACT · MTIL Order-I | Last accuracy: 85.60 -> 90.54 | Improved |
| PACT vs best prior baseline on MTIL | CLIP ViT-B/16 · MTIL Order-I | Last accuracy: +2.94 | Improved |
| use PACT in few-shot MTIL 16-shot | CLIP ViT-B/16 · FS-MTIL 16-shot | Last accuracy: 77.7 -> 84.4 | Improved |
| use PACT in CIL on CIFAR100 (10 steps) | CLIP ViT-B/16 · CIFAR100 CIL | Last accuracy: 79.33 -> 80.57 | Improved |
Closest papers by meaning
- Interference-Aware Continual Vision–Language Learning via Instance-Level Expert Routing (ECCV 2026 · similarity 0.90)
- Enhanced Continual Learning of Vision-Language Models with Model Fusion (ICLR 2026 · similarity 0.89)
- Boosting Continual Learning of Vision-Language Models via Mixture-of-Experts Adapters (CVPR 2024 · similarity 0.89)
- Beyond Disjoint Tasks: Towards More Natural Continual Learning for Vision-Language Models (ECCV 2026 · similarity 0.89)
- Overcoming Generic Knowledge Loss with Selective Parameter Update (CVPR 2024 · similarity 0.87)
- Synthetic Data is an Elegant GIFT for Continual Vision-Language Models (CVPR 2025 · similarity 0.87)
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.
