Regular Papers
Up to 8 pages excluding references and appendix
Machine Learning/CORE Unranked
CL4FMAgents @ NeurIPS 2026
NeurIPS 2026 Workshop on Continual Learning in the Era of Foundation Models and Embodied Agents
The CL4FMAgents @ NeurIPS 2026 workshop explores continual learning as a shared challenge between foundation models and embodied agents in dynamic, open-ended environments. It brings together researchers to address issues like catastrophic forgetting, self-evolution, online adaptation, and safety during lifelong learning, with a focus on cross-pollination between continual learning, LLM agents, robotics, and embodied intelligence.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Short Papers
Up to 4 pages excluding references and appendix
Research areas in scope
01
Continual Learning for Foundation Models
Continual and lifelong pre-training, post-training, and alignment of foundation models (LLMs, VLMs, and multimodal models)
02
LLM and Foundation-Model Agents
Continual learning, self-evolution, and long-term skill acquisition of LLM and foundation-model agentsAgent recursive self-improvement and open-ended self-evolution of agentic systemsAgent reinforcement learning for continual and lifelong adaptationLoop engineering and harness engineering for continually improving agents (scaffolding, tool use, feedback and self-refinement loops, and evaluation harnesses)
03
Embodied Agents and Robotics
Continual and lifelong learning for embodied agents, robotics, control, and world models
04
Architectures and Adaptation
Memory architectures, knowledge consolidation, model editing, and retrieval-augmented adaptationOnline, test-time, and streaming adaptation under distribution shift and non-stationarityCatastrophic forgetting, stability–plasticity trade-offs, and forward/backward transfer
05
Theoretical Foundations
Theory and foundations of continual learning: generalization, optimization, and scaling laws
06
Safety and Reliability
Safety, robustness, privacy, and reliability during continual updates and self-improvement
07
Evaluation and Applications
Benchmarks, evaluation protocols, and metrics for lifelong and continually learning systemsApplications and deployment: scientific discovery, healthcare, autonomous systems, and personalization
Policies worth checking twice
- All submissions are reviewed double-blind
- Submissions are non-archival
- Submissions must not re-present finalized work previously published at ML venues
- Each submission will receive at least three reviews
- Accepted papers will be presented in poster sessions, with selected papers featured as oral talks
- A Best Paper Award will be offered
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026