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

2026-02-11

AI Summary

Existing approaches to modeling human social behavior often treat heterogeneous behavioral dimensions in isolation, leading to high training costs and poor generalization. To address this, this work proposes Heterogeneous-Aware Relative Policy Optimization (HARPO), which, for the first time in reinforcement learning, explicitly models a balancing mechanism over heterogeneous behavioral data. HARPO modulates the advantage function to harmonize multi-task and multi-sample policy updates, enabling unified and stable social behavior modeling. Built upon this framework, we develop OmniSapiens-7B 2.0, a 7-billion-parameter foundation model for social behavior that achieves performance gains of 16.85% and 9.37% in multi-task and zero-shot settings, respectively, significantly outperforming existing methods and demonstrating consistently strong results across diverse behavioral tasks.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Cognitive Modeling & Cognitive Systems: Simulating Human Behavior
• Search and Optimization: Learning to Search

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• User Modeling, Personalization and Recommendation: Social recommender systems and personalization for the social Web
• Economics, Online Markets and Human Computation: Social networks and social learning

Abstract

To develop socially intelligent AI, existing approaches typically model human behavioral dimensions (e.g., affective, cognitive, or social attributes) in isolation. Although useful, task-specific modeling often increases training costs and limits generalization across behavioral settings. Recent reasoning RL methods facilitate training a single unified model across multiple behavioral tasks, but do not explicitly address learning across different heterogeneous behavioral data. To address this gap, we introduce Heterogeneity-Aware Relative Policy Optimization (HARPO), an RL method that balances leaning across heterogeneous tasks and samples. This is achieved by modulating advantages to ensure that no single task or sample carries disproportionate influence during policy optimization. Using HARPO, we develop and release Omnisapiens-7B 2.0, a foundation model for social behavior processing. Relative to existing behavioral foundation models, Omnisapiens-7B 2.0 achieves the strongest performance across behavioral tasks, with gains of up to +16.85% and +9.37% on multitask and held-out settings respectively, while producing more explicit and robust reasoning traces. We also validate HARPO against recent RL methods, where it achieves the most consistently strong performance across behavioral tasks.

Problem

• social behavior processing
• heterogeneous behavioral data
• foundation model
• multi-task learning
• generalization

Innovation

• Heterogeneity-Aware Relative Policy Optimization
• Foundation Model
• Social Behavior Processing
• Reinforcement Learning
• Multitask Generalization