Adding a penalty for ally deaths to distributional multi-agent Q-learning improves win rates on StarCraft II and driving benchmarks compared with six baseline algorithms.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Tackling Uncertainties in Multi-Agent Reinforcement Learning through Integration of Agent Termination Dynamics
Adding a penalty for ally deaths to distributional multi-agent Q-learning improves win rates on StarCraft II and driving benchmarks compared with six baseline algorithms.