MARS: Reinforcing Multi-Agent Reasoning of LLMs through Self-Play in Strategic Games
Paper
ā¢
2510.15414
ā¢
Published
ā¢
1
MARSHAL
MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs š Accepted by ICLR 2026
Note Note: This paper has been updated to v2 on arXiv. MARSHAL: Incentivizing Multi-Agent Reasoning via Self-Play with Strategic LLMs
-
nics-efc/MARSHAL-Generalist-Qwen3-4B
Text Generation ā¢ 4B ā¢ Updated ā¢ 8 -
nics-efc/MARSHAL-Generalist-Qwen3-8B
Text Generation ā¢ 8B ā¢ Updated ā¢ 9 -
nics-efc/MARSHAL-Tic-Tac-Toe-Qwen3-4B
Text Generation ā¢ 4B ā¢ Updated ā¢ 10 -
nics-efc/MARSHAL-Kuhn-Poker-Qwen3-4B
Text Generation ā¢ 4B ā¢ Updated ā¢ 7 -
nics-efc/MARSHAL-Mini-Hanabi-Qwen3-4B
Text Generation ā¢ 4B ā¢ Updated ā¢ 5