HumanLLM: Benchmarking and Improving LLM Anthropomorphism via Human Cognitive Patterns
This addresses the problem of improving persona simulation and role-playing agents for AI applications, though it appears incremental as it builds on existing LLM frameworks with a novel benchmarking approach.
The paper tackles the challenge of aligning LLMs with human cognitive patterns for authentic anthropomorphism, achieving strong human alignment (r=0.91) and showing that HumanLLM-8B outperforms Qwen3-32B on multi-pattern dynamics despite having 4x fewer parameters.
Large Language Models (LLMs) have demonstrated remarkable capabilities in reasoning and generation, serving as the foundation for advanced persona simulation and Role-Playing Language Agents (RPLAs). However, achieving authentic alignment with human cognitive and behavioral patterns remains a critical challenge for these agents. We present HumanLLM, a framework treating psychological patterns as interacting causal forces. We construct 244 patterns from ~12,000 academic papers and synthesize 11,359 scenarios where 2-5 patterns reinforce, conflict, or modulate each other, with multi-turn conversations expressing inner thoughts, actions, and dialogue. Our dual-level checklists evaluate both individual pattern fidelity and emergent multi-pattern dynamics, achieving strong human alignment (r=0.91) while revealing that holistic metrics conflate simulation accuracy with social desirability. HumanLLM-8B outperforms Qwen3-32B on multi-pattern dynamics despite 4x fewer parameters, demonstrating that authentic anthropomorphism requires cognitive modeling--simulating not just what humans do, but the psychological processes generating those behaviors.