第四屆終身學習智能體會議（CoLLAs）主題征稿中

?CoLLAs 2025 即將開辦，入選論文將發表在《機器學習研究會議錄》（PMLR），誠邀 AI、機器學習等相關領域研究人員投稿！

原標題：第四屆終身學習智能體會議（CoLLAs）主題征稿中
文章來源：AI科技評論
內容字數：6116字

CoLLAs 2025: Call for Papers on Lifelong Learning Agents

The Conference on Lifelong Learning Agents (CoLLAs) 2025,hosted by the University of Pennsylvania in Philadelphia,invites submissions from researchers in AI and machine learning. This annual conference focuses on advancing machine learning systems capable of continuous learning throughout their lifespan,exhibiting increasing knowledge and capability,and performing effectively in complex,non-stationary environments. Accepted papers will be published in the Proceedings of Machine Learning Research (PMLR).

1. Conference Overview

CoLLAs addresses the limitations of traditional machine learning,which often assumes independent and identically distributed (i.i.d.) data and focuses solely on single-task optimization. The conference aims to explore alternative paradigms that embrace continual learning in non-i.i.d. and non-stationary settings,mimicking the adaptive learning capabilities of human intelligence. Organized by researchers from leading institutions including Mila,Google DeepMind,Google Brain,and several prominent universities,CoLLAs boasts a prestigious advisory board comprised of renowned experts in the field.

2. CoLLAs 2025 Organization

CoLLAs 2025 will be the fourth edition of this conference. The General Chairs are Sarath Chandar (Mila) and Razvan Pascanu (Google DeepMind). The Program Chairs include Eric Eaton (University of Pennsylvania),Bing Liu (University of Illinois Chicago),Rupam Mahmood (University of Alberta),and Amal Rannen-Triki (Google DeepMind),supported by numerous Associate Program Chairs from academia and industry.

3. Topics and Call for Papers

The conference welcomes submissions on a wide range of topics related to continual learning,including (but not limited to):

1. Theoretical foundations of continual/lifelong learning
2. Continual learning paradigms (e.g.,class-incremental,task-incremental,domain-incremental,curriculum learning,active learning,federated learning,online learning,meta-learning,few-shot learning)
3. Challenges in non-stationary learning (e.g.,catastrophic forgetting,distribution shift,out-of-distribution generalization)
4. Continual reinforcement learning (e.g.,options,skill discovery,hierarchical RL,intrinsically motivated learning,multi-agent RL)
5. Continual learning in large language models (e.g.,in-context learning,pre-training,model editing,fine-tuning,adaptation)
6. Knowledge transfer (e.g.,transfer learning,multi-task learning,domain adaptation,simulation-to-reality,meta-learning)
7. Non-stationary optimization techniques
8. Streaming learning,on-device,and real-time learning
9. Open-world and open-ended learning
10. Neuroscience-inspired continual learning
11. Applications in control,robotics,healthcare,etc.
12. Datasets,benchmarks,evaluation metrics,and software libraries

Submissions exploring the intersection of continual learning with other fields (neuroscience,robotics,education) are especially encouraged. Submitted papers will be assessed based on novelty,technical rigor,and potential impact. Reproducibility is crucial,and authors are urged to make code and data publicly available.

4. Important Dates

Abstract Deadline: February 21,2025
Submission Deadline: February 26,2025
Reviews Released: April 7,2025
Author Rebuttals Due: April 15,2025
Notification of Decision: May 12,2025
Resubmissions Deadline: June 12,2025
Decision on Resubmissions: June 23,2025

For inquiries,contact contact@lifelong-ml.cc

聯系作者

文章來源：AI科技評論
作者微信：
作者簡介：雷峰網旗下AI新媒體。聚焦AI前沿研究，關注AI工程落地。