Unverified50% confidenceFactExact time
In multi-agent RL, the simultaneous updating of each agent's policy causes the environment (joint policy) faced by the other agents to change continuously, breaking the stationarity assumption of the Markov decision process in single-agent RL.
1
Sources
50%
Confidence
Long-term
Relevance
10/27/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedMulti-agent systems exhibit non-stationarity, as the optimal policy for each agent depends on the behavior of the other agents.75% similarVerifiedMulti-Agent Reinforcement Learning (MARL) suffers from non-stationarity: each agent's optimal policy depends on the continuously updating policies of other agents, causing the environment's transition distribution to change constantly.74% similarUnverified多智能体强化学习领域的非平稳性问题在多人世界模型中被放大，因为其他玩家的行为分布会随策略动态改变72% similarUnverifiedWhen the policies of other agents are fixed, the environment degenerates into a standard MDP for agent i, and is thus "individually stationary."72% similarUnverifiedA typical defect of multi-agent systems is that each agent can only observe local information, leading to information asymmetry problems where decisions are made based on an outdated world model.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1103310API
curl https://kongchang.com/api/v1/knowledge/claims/1103310MCP
get_claim(id=1103310)