MENTIS: Simulating Minds in World Models to Predict Human Decisions
mental-world-model · hf · 2026-08-03
While existing world models predict physical scene evolution, human behavior is driven by hidden mental states (beliefs, intents, social norms). To bridge this gap, researchers proposed Mental World Modeling (MWM), a theoretical framework making mental variables core components of a world model, maintaining a coupled physical-mental state.
- MENTIS Baseline: The team introduced MENTIS, a training-free, fully inspectable baseline that decomposes the process into state parsing, target-observation generation, action decomposition, and coupled physical-mental transitions.
- Key Findings: Evaluated on a dataset of situated decision scenarios (text, image, and sounding-video stories) with 8 modern LLM-based world models, explicitly modeling mental states proved essential for accurately predicting human decisions.
MWM represents the next stage of world modeling: moving from simulating physical scenes to simulating the minds acting within them.
More from Research
- Explained: How AI Companies Achieve 10x Faster Video Model Inference — haremlifegame · 2026-08-03
- Echoverse: Training Agents in Deep Environments Boosts 9B Model Success Rate to 67% — burny_tech · 2026-08-03
- EMNLP 2025 submissions near 10k, researcher says the system is broken — prajdabre · 2026-08-03
- Foundation Model Uses Rich Sleep Data to Predict Disease Risk and Survival — EricTopol · 2026-08-03
- TUM's SGTP: Real-Time Game-Theoretic Planning for Autonomous Racing — TUM-AVS · 2026-08-03
- NJU Proposes CSCR: Fixing Credit Assignment in Long-CoT RL — NJU · 2026-08-03