Beyond Large Language Models: A Mental World Model Framework for Mental Modeling, Simulation, Evolution, and Interpretability
arXiv cs.AI · 17 Oct 2026, 02:42 AEST
A new framework called the Mental World Model (MWM) aims to enhance artificial intelligence by simulating subjective cognition and mental states. The Multi-Branch Mental Evolution Architecture (MMEA) demonstrates significant improvements in mental state prediction and behavior modeling compared to existing methods.
Key facts
- Understanding and simulating how subjective cognition and mental states emerge, evolve, and shape behavior is essential for advancing artificial intelligence toward deeper social cognition and general intelligence.
- This paper introduces the Mental World Model (MWM) and its implementation, the Multi-Branch Mental Evolution Architecture (MMEA), which explicitly constructs subjective worlds, models interactions between cognition and mental states, and simulates continuous mental evolution shaped by historical experiences.
- On DynToM and RecToM, MMEA achieves 96.48% and 94.00% accuracy, respectively, outperforming leading LLMs and specialized ToM methods.
- The results demonstrate that MMEA provides an interpretable framework for continuous mental modeling, enabling AI to move beyond surface-level semantic representations to infer individuals' underlying mental states from linguistic information.
- MWM also provides a promising foundation for causal reasoning, personalized AI, sustained human-AI interaction, and future advances toward artificial general intelligence (AGI).
Read aloud
A new framework called the Mental World Model MWM aims to enhance artificial intelligence by simulating subjective cognition and mental states. The Multi-Branch Mental Evolution Architecture MMEA demonstrates significant improvements in mental state prediction and behavior modeling compared to existing methods. Understanding and simulating how subjective cognition and mental states emerge, evolve, and shape behavior is essential for advancing artificial intelligence toward deeper social cognition and general intelligence. This paper introduces the Mental World Model MWM and its implementation, the Multi-Branch Mental Evolution Architecture MMEA, which explicitly constructs subjective worlds, models interactions between cognition and mental states, and simulates continuous mental evolution shaped by historical experiences. On DynToM and RecToM, MMEA achieves ninety six point four eight percent and ninety four point zero zero percent accuracy, respectively, outperforming leading L L M s and specialized ToM methods. The results demonstrate that MMEA provides an interpretable framework for continuous mental modeling, enabling A I to move beyond surface-level semantic representations to infer individuals' underlying mental states from linguistic information. MWM also provides a promising foundation for causal reasoning, personalized A I, sustained human-A I interaction, and future advances toward artificial general intelligence AGI.
Sources in this cluster
- Beyond Large Language Models: A Mental World Model Framework for Mental Modeling, Simulation, Evolution, and Interpretability
arXiv cs.AI · 17 Oct 2026, 02:42 AEST