Deceive, Detect, and Disclose: Large Language Models Play Mini-Mafia
This work provides a benchmark for assessing interactive capabilities like deception and detection in LLMs, with implications for AI safety and multi-agent dynamics, though it is incremental in adapting an existing game framework.
The authors tackled the problem of evaluating social intelligence in large language models (LLMs) by introducing Mini-Mafia, a simplified four-player variant of the social deduction game Mafia, and found counterintuitive results such as smaller models outperforming larger ones in some cases.
Mafia is a social deduction game where informed mafia compete against uninformed townsfolk. Its asymmetry of information and reliance on theory-of-mind reasoning mirror real-world multi-agent scenarios, making it a useful testbed for evaluating the social intelligence of large language models (LLMs). To support a systematic study, we introduce Mini-Mafia: a simplified four-player variant with one mafioso, one detective, and two villagers. We set the mafioso to kill a villager and the detective to investigate the mafioso during the night, reducing the game to a single day phase of discussion and voting. This setup isolates three interactive capabilities through role-specific win conditions: the mafioso must deceive, the villagers must detect deception, and the detective must effectively disclose information. To measure these skills, we have LLMs play against each other, creating the Mini-Mafia Benchmark: a two-stage framework that first estimates win rates within fixed opponent configurations, then aggregates performance across them using standardized scoring. Built entirely from model interactions without external data, the benchmark evolves as new models are introduced, with each one serving both as a new opponent and as a subject of evaluation. Our experiments reveal counterintuitive results, including cases where smaller models outperform larger ones. Beyond benchmarking, Mini-Mafia enables quantitative study of emergent multi-agent dynamics such as name bias and last-speaker advantage. It also contributes to AI safety by generating training data for deception detectors and by tracking models' deception capabilities against human baselines.