When Reasoning Models Hurt Behavioral Simulation: A Solver-Sampler Mismatch in Multi-Agent LLM Negotiation

Build a benchmark suite for multi-agent social simulations that tests for the 'Solver-Sampler Mismatch.' Help researchers ensure their agents act like humans, not hyper-rational optimizers.

Suggested repo: simuLogic

"Stop your agents from being too smart to simulate human behavior."

Estimated effort: 45h