Post-training makes large language models less human-like

Marcel Binz, Elif Akata, Abdullah Almaatouq +76

Large language models (LLMs) are increasingly used as surrogates for human participants, but it remains unclear which models best capture human behavior and why. To address this, w…