ICML 2025 accepted papers on Continual learning
51 of the 3,342 papers accepted at ICML 2025 (1.5%) are on Continual learning, compared with 1.0% at ICML 2024. They fall into 4 topics.
See ICML 2025 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
Online and replay-based · 23 papers
- Activation by Interval-wise Dropout: A Simple Way to Prevent Neural Networks from Plasticity Loss (2 citations)
- FedSSI: Rehearsal-Free Continual Federated Learning with Synergistic Synaptic Intelligence (2 citations)
- Optimal Task Order for Continual Learning of Multiple Tasks (1 citation)
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (1 citation)
- Rethinking the Stability-Plasticity Trade-off in Continual Learning from an Architectural Perspective (1 citation)
- Where is the Truth? The Risk of Getting Confounded in a Continual World (1 citation)
- BECAME: Bayesian Continual Learning with Adaptive Model Merging
- Continual Reinforcement Learning by Planning with Online World Models
- Cut out and Replay: A Simple yet Versatile Strategy for Multi-Label Online Continual Learning
- Improving Continual Learning Performance and Efficiency with Auxiliary Classifiers
- Knowledge Retention in Continual Model-Based Reinforcement Learning
- Learning without Isolation: Pathway Protection for Continual Learning
and 11 more in the interactive map.
Continual fine-tuning of LLMs · 12 papers
- How Much Can We Forget about Data Contamination? (1 citation)
- LADA: Scalable Label-Specific CLIP Adapter for Continual Learning
- Large Continual Instruction Assistant
- Learning Dynamics in Continual Pre-Training for Large Language Models
- Proxy-FDA: Proxy-based Feature Distribution Alignment for Fine-tuning Vision Foundation Models without Forgetting
- SEFE: Superficial and Essential Forgetting Eliminator for Multimodal Continual Instruction Tuning
- TreeLoRA: Efficient Continual Learning via Layer-Wise LoRAs Guided by a Hierarchical Gradient-Similarity Tree
- Upweighting Easy Samples in Fine-Tuning Mitigates Forgetting
- Adaptive Localization of Knowledge Negation for Continual LLM Unlearning
- Dynamic Mixture of Curriculum LoRA Experts for Continual Multimodal Instruction Tuning
- Exploiting Presentative Feature Distributions for Parameter-Efficient Continual Learning of Large Language Models
- Retrieval-Augmented Language Model for Knowledge-aware Protein Encoding
Class-incremental · 10 papers
- A Selective Learning Method for Temporal Graph Continual Learning
- Autoencoder-Based Hybrid Replay for Class-Incremental Learning
- Continual Generalized Category Discovery: Learning and Forgetting from a Bayesian Perspective
- Demystifying Catastrophic Forgetting in Two-Stage Incremental Object Detector
- Knowledge Swapping via Learning and Unlearning
- L3A: Label-Augmented Analytic Adaptation for Multi-Label Class Incremental Learning
- Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning
- EmoGrowth: Incremental Multi-label Emotion Decoding with Augmented Emotional Relation Graph
- Semantic Shift Estimation via Dual-Projection and Classifier Reconstruction for Exemplar-Free Class-Incremental Learning
- Tensor Decomposition Based Memory-Efficient Incremental Learning
Domain and prompt-based · 6 papers
- Addressing Imbalanced Domain-Incremental Learning through Dual-Balance Collaborative Experts
- Componential Prompt-Knowledge Alignment for Domain Incremental Learning
- DPCore: Dynamic Prompt Coreset for Continual Test-Time Adaptation
- GCAL: Adapting Graph Models to Evolving Domain Shifts
- Ranked Entropy Minimization for Continual Test-Time Adaptation
- Scaling Laws for Forgetting during Finetuning with Pretraining Data Injection