MindZero: Learning Online Mental Reasoning With Zero Annotations</p>\n","updatedAt":"2026-06-02T23:11:46.244Z","author":{"_id":"64beb69801f1983a86a05de2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64beb69801f1983a86a05de2/tFyCoqZ6gT8NWkZfuncID.jpeg","fullname":"Chuanyang Jin","name":"Chuanyang-Jin","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":10,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.31505581736564636},"editors":["Chuanyang-Jin"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/64beb69801f1983a86a05de2/tFyCoqZ6gT8NWkZfuncID.jpeg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.00240","authors":[{"_id":"6a1f630ce292c1c78ecb1259","name":"Shunchi Zhang","hidden":false},{"_id":"6a1f630ce292c1c78ecb125a","name":"Jin Lu","hidden":false},{"_id":"6a1f630ce292c1c78ecb125b","name":"Chuanyang Jin","hidden":false},{"_id":"6a1f630ce292c1c78ecb125c","name":"Yichao Zhou","hidden":false},{"_id":"6a1f630ce292c1c78ecb125d","name":"Zhining Zhang","hidden":false},{"_id":"6a1f630ce292c1c78ecb125e","name":"Tianmin Shu","hidden":false}],"publishedAt":"2026-05-29T00:00:00.000Z","submittedOnDailyAt":"2026-06-02T00:00:00.000Z","title":"MindZero: Learning Online Mental Reasoning With Zero Annotations","submittedOnDailyBy":{"_id":"64beb69801f1983a86a05de2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64beb69801f1983a86a05de2/tFyCoqZ6gT8NWkZfuncID.jpeg","isPro":false,"fullname":"Chuanyang Jin","user":"Chuanyang-Jin","type":"user","name":"Chuanyang-Jin"},"summary":"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.","upvotes":2,"discussionId":"6a1f630ce292c1c78ecb125f","projectPage":"https://scai.cs.jhu.edu/MindZero/","githubRepo":"https://github.com/SCAI-JHU/MindZero","githubRepoAddedBy":"user","ai_summary":"MindZero presents a self-supervised reinforcement learning framework that enables multimodal large language models to perform efficient and robust online mental reasoning without requiring explicit mental state annotations.","ai_keywords":["Theory of Mind","reinforcement learning","multimodal large language models","model-based reasoning","self-supervised learning","mental state hypotheses","planner","gridworld","household domains"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","githubStars":6,"organization":{"_id":"653945b47ba797097a7b4eab","name":"JohnsHopkins","fullname":"Johns Hopkins University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/653944e58e687a41625a4694/qqHzBOarppVrUuZbbjqwh.png"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"64beb69801f1983a86a05de2","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64beb69801f1983a86a05de2/tFyCoqZ6gT8NWkZfuncID.jpeg","isPro":false,"fullname":"Chuanyang Jin","user":"Chuanyang-Jin","type":"user"},{"_id":"66cbea552400073af31bdf7f","avatarUrl":"/avatars/5de3c4da60c6097b6bd4ce245113e1c2.svg","isPro":false,"fullname":"Social Cognitive AI Lab at JHU","user":"SCAI-JHU","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"653945b47ba797097a7b4eab","name":"JohnsHopkins","fullname":"Johns Hopkins University","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/653944e58e687a41625a4694/qqHzBOarppVrUuZbbjqwh.png"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.00240.md"}">
MindZero: Learning Online Mental Reasoning With Zero Annotations
Abstract
MindZero presents a self-supervised reinforcement learning framework that enables multimodal large language models to perform efficient and robust online mental reasoning without requiring explicit mental state annotations.
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.
Community
MindZero: Learning Online Mental Reasoning With Zero Annotations
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2606.00240 in a model README.md to link it from this page.
Cite arxiv.org/abs/2606.00240 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2606.00240 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.