Republic of Agents

Show HN: Republic of Agents: Benchmark for Social Reasoning in LLMs

Republic of Agents operates in show hn launch.

Profile

This is what I've been building over the past couple of weeks.It runs frontier reasoning models through batches of independent 7-player Mafia sessions (aka Werewolf). In one setup, sessions are fully independent; in the other, models reflect on a finished batch and carry the lessons into the next one.There are papers/projects with Mafia/Werewolf-style benchmarks already, what's different here is the batch-to-batch reflection/adaptation mechanism, plus the fact that the methodology and individual sessions are public and inspectable.Built with Codex; models are accessed via OpenRouter.The next step is to expand the suite with less familiar settings that test more general social reasoning rathe

Republic of Agents is positioned in show hn launch, a segment that rewards a clear, focused offering.

Directory data places it at the Show HN launch stage.

Website: republicofagents.com