First provide a summary of the paper, and then address the following criteria: Quality, clarity, originality and significance. The paper considers multi-agent systems where a voting rule is used to aggregate decisions of individual agents to solve a problem. One can consider the agents as sampling from a noisy distribution centered around the true ranking of the alternatives at each state. They show that if the agents are copies of each other, in the sense that the agents produce samples (preferences for alternatives) from the same distribution then it is likely that system votes for a sub-optimal alternative. On the other hand, if we have many agents that have sufficient diversity then the system is likely to vote for the optimal alternatives in almost every state.

Neural Information Processing Systems http://nips.cc/

Temporal prediction is critical for making intelligent and robust decisions in complex dynamic environments. Motion prediction needs to model the inherently uncertain future which often contains multiple potential outcomes, due to multi-agent interactions and the latent goals of others.

It also performs better empirically.

Diplomacy is a seven-player game where players attempt to acquire a majority of supply centers across Europe. To acquire supply centers, players can coordinate their units with other players through dialogue or signaling.

V alue factorisation is a useful technique for multi-agent reinforcement learning (MARL) in global reward game, however, its underlying mechanism is not yet fully understood. This paper studies a theoretical framework for value factorisation with interpretability via Shapley value theory.

Language is an interface to the outside world.