We aim to systematically investigate the key question: What can multimodal Chain-of-Thought reasoning do, and where and why does it fall short?</p>\n","updatedAt":"2026-06-25T03:00:13.434Z","author":{"_id":"643379416c6ecd58798421b3","avatarUrl":"/avatars/831db7eab2663abc33b176cf386b02f2.svg","fullname":"Zhuoran Jin","name":"jinzhuoran","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":11,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9094933867454529},"editors":["jinzhuoran"],"editorAvatarUrls":["/avatars/831db7eab2663abc33b176cf386b02f2.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.22565","authors":[{"_id":"6a3c991ef3facdb67e9ff0f1","name":"Zhuoran Jin","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f2","name":"Kejian Zhu","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f3","name":"Hongbang Yuan","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f4","name":"Yupu Hao","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f5","name":"Pengfei Cao","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f6","name":"Yubo Chen","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f7","name":"Kang Liu","hidden":false},{"_id":"6a3c991ef3facdb67e9ff0f8","name":"Jun Zhao","hidden":false}],"publishedAt":"2026-06-21T00:00:00.000Z","submittedOnDailyAt":"2026-06-25T00:00:00.000Z","title":"Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do","submittedOnDailyBy":{"_id":"643379416c6ecd58798421b3","avatarUrl":"/avatars/831db7eab2663abc33b176cf386b02f2.svg","isPro":false,"fullname":"Zhuoran Jin","user":"jinzhuoran","type":"user","name":"jinzhuoran"},"summary":"Chain-of-Thought (CoT) has become a standard method for improving reasoning capabilities in large language models (LLMs) by eliciting step-by-step thinking, but its effectiveness in multimodal tasks remains unclear. In this paper, we aim to systematically investigate the key question: What can multimodal Chain-of-Thought reasoning do, and where and why does it fall short? To this end, we evaluate 12 multimodal tasks across perception and reasoning categories using both 14 non-reasoning models and 8 reasoning models. Our analysis reveals several important findings: (1) CoT is not a free lunch and should be used selectively depending on the specific requirements of each task. For perception tasks, CoT can lead to undesirable side effects, such as reduced performance in visual grounding and object counting. In contrast, it proves effective for reasoning tasks involving mathematical, scientific, and multi-image reasoning; (2) Compared to original models, existing open-source multimodal reasoning models often yield only marginal overall improvements, possibly due to an overemphasis on mathematical reasoning at the expense of broader capabilities; (3) Visual reasoning remains a key bottleneck for current multimodal CoT, as models exhibit a Look Light, Think Heavy pattern where verbal reflection rises and falls during reasoning, whereas visual reflection consistently diminishes. These findings suggest that while multimodal CoT handles verbal reflection relatively well, it lacks the ability to maintain deep visual introspection throughout the reasoning process.","upvotes":6,"discussionId":"6a3c991ef3facdb67e9ff0f9","ai_summary":"Multimodal Chain-of-Thought reasoning shows selective effectiveness across different tasks, with limitations in maintaining visual introspection during reasoning processes.","ai_keywords":["Chain-of-Thought","multimodal tasks","large language models","visual grounding","object counting","mathematical reasoning","scientific reasoning","multi-image reasoning","visual reasoning","Look Light Think Heavy pattern"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct","organization":{"_id":"640a887796aae649741a586f","name":"CASIA","fullname":"Chinese Academic of Science Institute of Automation","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1678411888885-6388984e8a5dbe2f3dc5afee.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"643379416c6ecd58798421b3","avatarUrl":"/avatars/831db7eab2663abc33b176cf386b02f2.svg","isPro":false,"fullname":"Zhuoran Jin","user":"jinzhuoran","type":"user"},{"_id":"6307612bfd79b417f1bc3fa3","avatarUrl":"/avatars/e86ed202106c43d5ba65bc3ff1f0c1fd.svg","isPro":false,"fullname":"ricky_33","user":"ricky333","type":"user"},{"_id":"654f3e104c8874c64d43aafa","avatarUrl":"/avatars/00de263f98a81c52cdb321fb11b16c06.svg","isPro":false,"fullname":"You Li","user":"Michael4933","type":"user"},{"_id":"683c8735a038b00c3cfa3c84","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/SZQBb7cMnW4NAuzPNaOco.png","isPro":false,"fullname":"Zixuan Cao","user":"MagicPenguin233","type":"user"},{"_id":"64b89dfa6a68a9a715df407e","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64b89dfa6a68a9a715df407e/FpBAdClhr-oVAv11Bjwjs.jpeg","isPro":false,"fullname":"Jiachun Li","user":"Septzzz","type":"user"},{"_id":"66935bdc5489e4f73c76bc7b","avatarUrl":"/avatars/129d1e86bbaf764b507501f4feb177db.svg","isPro":false,"fullname":"Abidoye Aanuoluwapo","user":"Aanuoluwapo65","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"640a887796aae649741a586f","name":"CASIA","fullname":"Chinese Academic of Science Institute of Automation","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/1678411888885-6388984e8a5dbe2f3dc5afee.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.22565.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2606.22565 in a model README.md to link it from this page.

Cite arxiv.org/abs/2606.22565 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2606.22565 in a Space README.md to link it from this page.

Add this paper to a collection to link it from this page.