OmniSapiens: A Foundation Model for Social Behavior Processing via Heterogeneity-Aware Relative Policy Optimization

OmniSapiens-7B 2.0 is a newly released foundation model designed for processing social behavior, leveraging Heterogeneity-Aware Relative Policy Optimization to effectively learn from diverse and imbalanced behavioral data. This model achieves superior performance across 10 behavioral tasks and five held-out benchmarks, with improvements of up to +12.02% and +9.37%, respectively, while providing more consistent and interpretable reasoning traces. This advancement is significant for practitioners in AI, as it enhances the ability to develop socially intelligent systems capable of adapting to varied human behaviors and contexts.