Social-R1: Towards Human-like Social Reasoning in LLMs

2026-03-10

AI Summary

This work addresses the limitations of large language models in social reasoning, particularly their difficulty in accurately interpreting social cues and mental states. To overcome this, the authors propose Social-R1, a reinforcement learning framework that employs trajectory-level alignment to apply multidimensional rewards throughout the entire reasoning process—supervising logical structure, information density, and cognitive plausibility—thereby moving beyond conventional paradigms that optimize only the final output. Additionally, they introduce ToMBench-Hard, an adversarial evaluation benchmark designed to comprehensively assess social cognition capabilities. Experimental results demonstrate that the trained 4B-parameter model outperforms larger-scale counterparts across eight diverse benchmarks, exhibiting both high efficiency and robustness in social reasoning tasks.

Technology Category

• Cognitive Modeling & Cognitive Systems: Social Cognition And Interaction
• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

While large language models demonstrate remarkable capabilities across numerous domains, social intelligence - the capacity to perceive social cues, infer mental states, and generate appropriate responses - remains a critical challenge, particularly for enabling effective human-AI collaboration and developing AI that truly serves human needs. Current models often rely on superficial patterns rather than genuine social reasoning. We argue that cultivating human-like social intelligence requires training with challenging cases that resist shortcut solutions. To this end, we introduce ToMBench-Hard, an adversarial benchmark designed to provide hard training examples for social reasoning. Building on this, we propose Social-R1, a reinforcement learning framework that aligns model reasoning with human cognition through multi-dimensional rewards. Unlike outcome-based RL, Social-R1 supervises the entire reasoning process, enforcing structural alignment, logical integrity, and information density. Results show that our approach enables a 4B parameter model to surpass much larger counterparts and generalize robustly across eight diverse benchmarks. These findings demonstrate that challenging training cases with trajectory-level alignment offer a path toward efficient and reliable social intelligence.

Problem

• social intelligence
• social reasoning
• human-AI collaboration
• mental state inference
• social cues

Innovation

• Social Reasoning
• Reinforcement Learning
• Process Supervision
• Adversarial Benchmark
• Human-AI Alignment