Researchers propose a fine-tuned multi-agent framework for detecting OCEAN personality traits from life narratives, addressing inconsistencies in single-model LLM inferences caused by pretraining biases. Sub-agents conditioned through masked language modeling and psychometric supervision adopt high, low, or neutral trait perspectives. A judge LLM aggregates outputs for final predictions. The framework was evaluated through quantitative and qualitative experiments including baselines and ablations.
No score is assigned. Sources and their independence are shown in the citation chain below.