Tonghan Wang

ROMA. Codes accompanying the paper "ROMA: Multi-Agent Reinforcement Learning with Emergent Roles" (ICML 2020 https://arxiv.org/abs/2003.08039)

NDQ. Codes accompanying the paper "Learning Nearly Decomposable Value Functions with Communication Minimization" (ICLR 2020)

RODE. Codes accompanying the paper "RODE: Learning Roles to Decompose Multi-Agent Tasks (ICLR 2021, https://arxiv.org/abs/2010.01523). RODE is a scalable role-based multi-agent learning method which effectively discovers roles based on joint action space decomposition according to action effects, establishing a new state of the art on the StarCraft multi-agent benchmark.

DOP. Codes accompanying the paper "DOP: Off-Policy Multi-Agent Decomposed Policy Gradients" (ICLR 2021, https://arxiv.org/abs/2007.12322)

EITI-EDTI. Codes accompanying the paper "Influence-Based Multi-Agent Exploration" (ICLR 2020 spotlight)

CASEC-MACO-benchmark. Codes accompanying the paper "Context-Aware Sparse Deep Coordination Graphs (https://arxiv.org/abs/2106.02886).

BundleFlow. We propose the first deep menu-based method for single-bidder combinatorial auctions that is dominant-strategy incentive compatible and revenue-optimizing, by adapting continuous flow matching for optimization problems.

Homophily-MARL. Code for "A Closer Look at Cooperation Emergence via Multi-Agent RL"

tonghanwang.github.io. CSS