AAAI 2025 accepted papers on Continual learning
49 of the 3,486 papers accepted at AAAI 2025 (1.4%) are on Continual learning, compared with 1.7% at AAAI 2024. They fall into 4 topics.
See AAAI 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 · 19 papers
- Rethinking and Improving Student Learning and Forgetting Processes for Attention based Knowledge Tracing Models (17 citations)
- MalCL: Leveraging GAN-Based Generative Replay to Combat Catastrophic Forgetting in Malware Classification (9 citations)
- MultiSFL: Towards Accurate Split Federated Learning via Multi-Model Aggregation and Knowledge Replay (9 citations)
- Region-Based Optimization in Continual Learning for Audio Deepfake Detection (7 citations)
- ConSense: Continually Sensing Human Activity with WiFi via Growing and Picking (6 citations)
- TinySubNets: An Efficient and Low Capacity Continual Learning Strategy (6 citations)
- Exploring Rationale Learning for Continual Graph Learning (4 citations)
- Few-Shot, No Problem: Descriptive Continual Relation Extraction (3 citations)
- Dynamic Expansion Diffusion Learning for Lifelong Generative Modelling (2 citations)
- On Local Overfitting and Forgetting in Deep Neural Networks (2 citations)
- Pareto Continual Learning: Preference-Conditioned Learning and Adaption for Dynamic Stability-Plasticity Trade-off (2 citations)
- Kolmogorov-Arnold Networks Still Catastrophically Forget but Differently from MLP (1 citation)
and 7 more in the interactive map.
Class-incremental · 16 papers
- MOS: Model Surgery for Pre-Trained Model-Based Class-Incremental Learning (15 citations)
- Adaptive Prototype Replay for Class Incremental Semantic Segmentation (10 citations)
- Maintaining Fairness in Logit-based Knowledge Distillation for Class-Incremental Learning (8 citations)
- Adaptive Decision Boundary for Few-Shot Class-Incremental Learning (6 citations)
- Pseudo Informative Episode Construction for Few-Shot Class-Incremental Learning (5 citations)
- Specifying What You Know or Not for Multi-Label Class-Incremental Learning (5 citations)
- Continual Learning Using a Kernel-Based Method Over Foundation Models (3 citations)
- LAGD: Local Topological-Alignment and Global Semantic-Deconstruction for Incremental 3D Semantic Segmentation (2 citations)
- Prototypical Replay with Old-class Focusing Knowledge Distillation for Incremental Named Entity Recognition (2 citations)
- AAKR: Adversarial Attack-based Knowledge Retention for Continual Semantic Segmentation (1 citation)
- Compress to One Point: Neural Collapse for Pre-Trained Model-Based Class-Incremental Learning (1 citation)
- DCA: Dividing and Conquering Amnesia in Incremental Object Detection (1 citation)
and 4 more in the interactive map.
Domain and prompt-based · 7 papers
- DASK: Distribution Rehearsing via Adaptive Style Kernel Learning for Exemplar-Free Lifelong Person Re-Identification (9 citations)
- DualCP: Rehearsal-Free Domain-Incremental Learning via Dual-Level Concept Prototype (5 citations)
- Adaptive Prompting for Continual Relation Extraction: A Within-Task Variance Perspective (3 citations)
- Few-Shot Audio-Visual Class-Incremental Learning with Temporal Prompting and Regularization (1 citation)
- FR²Seg: Continual Segmentation Across Multiple Sites via Fourier Style Replay and Adaptive Consistency Regularization (1 citation)
- CAPrompt: Cyclic Prompt Aggregation for Pre-Trained Model Based Class Incremental Learning
- ConDo: Continual Domain Expansion for Absolute Pose Regression
Continual fine-tuning of LLMs · 7 papers
- MoE-LPR: Multilingual Extension of Large Language Models Through Mixture-of-Experts with Language Priors Routing (9 citations)
- CMT: A Memory Compression Method for Continual Knowledge Learning of Large Language Models (4 citations)
- Distribution-Level Feature Distancing for Machine Unlearning: Towards a Better Trade-off Between Model Utility and Forgetting (2 citations)
- Filling Memory Gaps: Enhancing Continual Semantic Parsing via SQL Syntax Variance-Guided LLMs Without Real Data Replay (1 citation)
- Walking the Web of Concept-Class Relationships in Incrementally Trained Interpretable Models
- Toward Efficient Data-Free Unlearning
- Training Consistent Mixture-of-Experts-Based Prompt Generator for Continual Learning