Gathering

The 'Gathering' dataset/benchmark contains scenarios for evaluating multi-agent Sequential Social Dilemmas (SSDs) and is used to assess the performance of AI agents in cooperative decision-making tasks.

Papers using Gathering (2)

Discovering Cooperative Pipelines: Autoresearch for Sequential Social Dilemmas2026Beyond Scalar Rewards: Dense Feedback for LLM Policy Synthesis in Sequential Social Dilemmas2026