# Socio-Foundation: A Model for Generalizable Individual Behavior Simulation via Hierarchical Capability Distillation

arXiv: https://arxiv.org/abs/2610.08967 · HTML: https://mlsanity.com/paper/2610.08967/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 60 references that are in the map.
- First posted on arXiv 6 Oct 2026.

## Placed in

- Knowledge distillation

## Builds on (most cited first, 25 of 25)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes](https://mlsanity.com/paper/2306.13649/) (2023)
- [LLMs Get Lost In Multi-Turn Conversation](https://mlsanity.com/paper/2505.06120/) (2025)
- [Scaling Synthetic Data Creation with 1,000,000,000 Personas](https://mlsanity.com/paper/2406.20094/) (2024)
- [LLM Agents Grounded in Self-Reports Enable General-Purpose Simulation of Individuals](https://mlsanity.com/paper/2411.10109/) (2024)
- [From Individual to Society: A Survey on Social Simulation Driven by Large Language Model-based Agents](https://mlsanity.com/paper/2412.03563/) (2024)
- [MOPD: Multi-Teacher On-Policy Distillation for Capability Integration in LLM Post-Training](https://mlsanity.com/paper/2606.30406/) (2026)
- [Unveiling the Truth and Facilitating Change: Towards Agent-based Large-scale Social Movement Simulation](https://mlsanity.com/paper/2402.16333/) (2024)
- [Flipping the Dialogue: Training and Evaluating User Language Models](https://mlsanity.com/paper/2510.06552/) (2025)
- [Mind the Sim2Real Gap in User Simulation for Agentic Tasks](https://mlsanity.com/paper/2603.11245/) (2026)
- [Finetuning LLMs for Human Behavior Prediction in Social Science Experiments](https://mlsanity.com/paper/2509.05830/) (2025)
- [HumanLM: Simulating Users with State Alignment Beats Response Imitation](https://mlsanity.com/paper/2603.03303/) (2026)
- [Centaur: a foundation model of human cognition](https://mlsanity.com/paper/2410.20268/) (2024)
- [Learning to Make MISTAKEs: Modeling Incorrect Student Thinking And Key Errors](https://mlsanity.com/paper/2510.11502/) (2025)
- [LoRI: Reducing Cross-Task Interference in Multi-Task Low-Rank Adaptation](https://mlsanity.com/paper/2504.07448/) (2025)
- [Customer-R1: Personalized Simulation of Human Behaviors via RL-based LLM Agent in Online Shopping](https://mlsanity.com/paper/2510.07230/) (2025)
- [OdysSim: Building Foundation Models for Human Behavior Simulation](https://mlsanity.com/paper/2606.14199/) (2026)
- [Be.FM: Open Foundation Models for Human Behavior](https://mlsanity.com/paper/2505.23058/) (2025)
- [HumanLLM: Towards Personalized Understanding and Simulation of Human Nature](https://mlsanity.com/paper/2601.15793/) (2026)
- [MORPHEUS: Modeling Role from Personalized Dialogue History by Exploring and Utilizing Latent Space](https://mlsanity.com/paper/2407.02345/) (2024)
- [Social-R1: Towards Human-like Social Reasoning in LLMs](https://mlsanity.com/paper/2603.09249/) (2026)
- [CURP: Codebook-based Continuous User Representation for Personalized Generation with LLMs](https://mlsanity.com/paper/2602.00742/) (2026)
- [Consolidating RLVR Capabilities Across Domains: A Deep Dive into Fusion Paradigms](https://mlsanity.com/paper/2608.27409/) (2026)
- [BehaveGPT: A Foundation Model for Large-scale User Behavior Modeling](https://mlsanity.com/paper/2505.17631/) (2025)
- [Beyond Isolated Behaviors: Hierarchical User Modeling for LLM Personalization](https://mlsanity.com/paper/2606.02300/) (2026)

## Abstract (arXiv)

Simulating individual behavior requires large language models (LLMs) to preserve persona traits while adapting to dynamic social contexts. However, general-purpose LLMs often flatten distinct personas, while task-specific tuning suffers from fragmentation and generalization. To overcome these challenges, we organize individual simulation into the FONTS Taxonomy, comprising five complementary capability dimensions: persona fidelity (F), outcome realization (O), behavioral naturalness (N), trajectory coherence (T), and social grounding (S). Grounded in this taxonomy, we curate a standardized training corpus library of approximately 10 million instances across 14 representative datasets and present Socio-Foundation. Socio-Foundation decouples specialization from integration via a three-stage pipeline: learning task experts via DAPO, consolidating them into capability experts via off-policy distillation, and unifying them via multi-teacher on-policy distillation (MOPD). We also establish IndiEval, consolidating 29 metrics across the FONTS dimensions. Experiments show that Socio-Foundation outperforms its textitQwen3-8B base by 11.0 points and approaches frontier models such as textitGLM-5.2, with ablations and out-of-distribution evaluations further demonstrating the effectiveness and generalization of our model.