Social-R1: Towards Human-like Social Reasoning in LLMs
This addresses the problem of enabling effective human-AI collaboration and developing AI that serves human needs by improving social reasoning, though it is incremental as it builds on existing reinforcement learning methods.
The paper tackled the challenge of social intelligence in large language models by introducing ToMBench-Hard, an adversarial benchmark for hard training examples, and Social-R1, a reinforcement learning framework that aligns reasoning with human cognition. The result showed that a 4B parameter model surpassed larger counterparts and generalized robustly across eight benchmarks.
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.