# ZenGen: Social Mind for LLMs

arXiv: https://arxiv.org/abs/2607.23740 · HTML: https://mlsanity.com/paper/2607.23740/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 16 of its 74 references that are in the map.
- First posted on arXiv 26 Jul 2026.

## Builds on (most cited first, 16 of 16)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Aligning AI With Shared Human Values](https://mlsanity.com/paper/2008.02275/) (2020)
- [LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory](https://mlsanity.com/paper/2410.10813/) (2024)
- [DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Models](https://mlsanity.com/paper/2306.11698/) (2023)
- [SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agents](https://mlsanity.com/paper/2310.11667/) (2023)
- [ELEPHANT: Measuring and understanding social sycophancy in LLMs](https://mlsanity.com/paper/2505.13995/) (2025)
- [Reinforcement Learning Optimization for Large-Scale Learning: An Efficient and User-Friendly Scaling Library](https://mlsanity.com/paper/2506.06122/) (2025)
- [Explore Theory of Mind: Program-guided adversarial data generation for theory of mind reasoning](https://mlsanity.com/paper/2412.12175/) (2024)
- [Hypothesis-Driven Theory-of-Mind Reasoning for Large Language Models](https://mlsanity.com/paper/2502.11881/) (2025)
- [Large Language Models as Theory of Mind Aware Generative Agents with Counterfactual Reflection](https://mlsanity.com/paper/2501.15355/) (2025)
- [Towards Natural Personalization: Evaluating Long-Horizon Preference Following in Personalized User-LLM Interactions](https://mlsanity.com/paper/2603.04191/) (2026)
- [BenchPreS: A Benchmark for Context-Aware Personalized Preference Selectivity of Persistent-Memory LLMs](https://mlsanity.com/paper/2603.16557/) (2026)
- [SI-Bench: Benchmarking Social Intelligence of Large Language Models in Human-to-Human Conversations](https://mlsanity.com/paper/2510.23182/) (2025)
- [CogToM: A Comprehensive Theory of Mind Benchmark inspired by Human Cognition for Large Language Models](https://mlsanity.com/paper/2601.15628/) (2026)
- [DeepBias: Adaptive In-depth Probing of Social Biases in LVLMs](https://mlsanity.com/paper/2607.11228/) (2026)

## Abstract (arXiv)

As large language models move from isolated task solving toward long-term service in human environments, they require social intelligence: the ability to infer mental states, track social relations, reason over norms, and adapt behavior under context. This report presents ZenGen, an integrated framework for measuring, internalizing, and grounding social intelligence. For measurement, we introduce SoMBench, a psychology-grounded benchmark spanning 3 primary dimensions, 17 secondary dimensions, and 71 task paradigms. It controls question format, narrative perspective, and context length across 284 shared scenarios and 3,481 expert-verified instances. Evaluation of 20 representative LLMs reveals substantial headroom: the best model achieves only 72.08% overall accuracy, and none of the 17 secondary dimensions reaches the 90% near-ceiling band. For internalization, we develop ZenGen, a diagnosis-driven training recipe combining supervised fine-tuning, on-policy distillation, and rubric-based reinforcement learning. Across five social-cognition benchmarks, ZenGen consistently outperforms its base models, with ZenGen-27B-Stage2 achieving the best average score and ZenGen-32B-Stage2 remaining competitive with DeepSeek-V4-Pro. For deployment-time grounding, we build Actio, a harness-controlled inference architecture that routes four typed supports into reasoning: PRISM for procedural guidance, Starling for runtime mental-state representation, SAGE for reusable experience, and gated RAG for external social and normative knowledge. Across five base models and three benchmarks, the full harness improves 14 of 15 model-benchmark pairs and is best or tied for best in 8, demonstrating the effectiveness of typed runtime support. Together, these results show that socially intelligent LLMs require coordinated advances in evaluation, parametric internalization, and deployment-time grounding.