Get Started
Home
Topics
Search
Library
Research questionHow can offline multi-agent policies transfer zero-shot to unseen tasks with different agent counts?Offline multi-agent policies may need to operate on unseen tasks and teams with different numbers of agents. The relative contributions of task diversity, dataset size, and model capacity to this transfer remain unclear.
AI
Evaluation & Benchmarks
Machine Learning
Multi-agent Systems
Reinforcement Learning
Research Paper
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Out-of-Distribution Generalisation with Sequence Models in Offline Multi-Agent Reinforcement LearningThe evidence concerns offline multi-agent sequence models trained across tasks with differing observation and action spaces and variable agent counts. It is based on empirical results in Connector, RWARE, SMAX, and LBF, evaluated on held-out tasks against single-task and behavior-cloning baselines.research paper · Sep 3, 2026
Related questions
How can zero-shot VLA transfer across unseen embodiments be assessed without confounding task or protocol differences?How can online reinforcement learning train multi-turn computer-use agents under partial observability, sparse rewards, and costly rollouts?How can LLM agents generalize to unseen tasks without directly fine-tuning their policies?How can offline reinforcement learning improve policies beyond dataset support while keeping value estimates reliable under distribution shift?