Didier Schwab

LLM benchmarks are missing a critical ingredient: social science data, which could significantly improve model generalization and robustness across a wide range of disciplines.