Search papers, labs, and topics across Lattice.
This paper introduces MindZero, a self-supervised reinforcement learning framework that enables multimodal large language models (MLLMs) to perform efficient online mental reasoning without requiring explicit annotations of mental states. By rewarding the generation of mental state hypotheses that align with observed actions, MindZero effectively integrates model-based reasoning into a single-pass inference process. The results demonstrate that MindZero significantly outperforms traditional model-based methods in both accuracy and efficiency across various mental reasoning and AI assistance tasks.
MindZero reveals that mental reasoning can be learned as a self-supervised skill, dramatically enhancing the performance of MLLMs in real-time AI assistance tasks.
Effective real-world assistance requires AI agents with robust Theory of Mind (ToM): inferring human mental states from their behavior. Despite recent advances, several key challenges remain, including (1) online inference with robust uncertainty updates over multiple hypotheses; (2) efficient reasoning suitable for real-time assistance; and (3) the lack of ground-truth mental state annotations in real-world domains. We address these challenges by introducing MindZero, a self-supervised reinforcement learning framework that trains multimodal large language models (MLLMs) for efficient and robust online mental reasoning. During training, the model is rewarded for generating mental state hypotheses that maximize the likelihood of observed actions estimated by a planner, similar to model-based ToM reasoning. This method thus eliminates the need for explicit mental state annotations. After training, MindZero internalizes model-based reasoning into fast single-pass inference. We evaluate MindZero against baselines across challenging mental reasoning and AI assistance tasks in gridworld and household domains. We found that LLMs alone are insufficient; model-based methods improve accuracy but are slow, costly, and limited by backbone MLLM capacity. In contrast, MindZero enhances MLLMs'intrinsic ToM ability and significantly outperforms model-based methods in both accuracy and efficiency, showing that mental reasoning can be effectively learned as a self-supervised skill.
