<a href=\"https://cdn-uploads.huggingface.co/production/uploads/659cf9791c8b66637e3de72d/I87T4oblWRUHVAHEJSFqh.png\" rel=\"nofollow\"><img src=\"https://cdn-uploads.huggingface.co/production/uploads/659cf9791c8b66637e3de72d/I87T4oblWRUHVAHEJSFqh.png\" alt=\"image\"></a></p>\n","updatedAt":"2026-08-14T06:32:20.175Z","author":{"_id":"659cf9791c8b66637e3de72d","avatarUrl":"/avatars/7d26710f687be9444796980662614f16.svg","fullname":"zhiqin yang","name":"visity","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.5735602974891663},"editors":["visity"],"editorAvatarUrls":["/avatars/7d26710f687be9444796980662614f16.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.07193","authors":[{"_id":"6a7eb64542823931a1f17799","name":"Zhen Liu","hidden":false},{"_id":"6a7eb64542823931a1f1779a","name":"Wenli Huang","hidden":false},{"_id":"6a7eb64542823931a1f1779b","name":"Wei Song","hidden":false},{"_id":"6a7eb64542823931a1f1779c","name":"Yuhan Liu","hidden":false},{"_id":"6a7eb64542823931a1f1779d","user":{"_id":"659cf9791c8b66637e3de72d","avatarUrl":"/avatars/7d26710f687be9444796980662614f16.svg","isPro":false,"fullname":"zhiqin yang","user":"visity","type":"user","name":"visity"},"name":"Zhiqin Yang","status":"claimed_verified","statusLastChangedAt":"2026-08-14T08:45:04.794Z","hidden":false},{"_id":"6a7eb64542823931a1f1779e","name":"Jingwen Fu","hidden":false}],"publishedAt":"2026-08-07T00:00:00.000Z","submittedOnDailyAt":"2026-08-14T00:00:00.000Z","title":"An AI4AI Framework for Visual Token Pruning","submittedOnDailyBy":{"_id":"659cf9791c8b66637e3de72d","avatarUrl":"/avatars/7d26710f687be9444796980662614f16.svg","isPro":false,"fullname":"zhiqin yang","user":"visity","type":"user","name":"visity"},"summary":"Visual-token pruning can substantially reduce the inference cost of multimodal large language models (MLLMs), yet existing methods largely rely on fixed, handcrafted heuristics and costly expert trial and error. As pruning objectives, budgets, and model architectures diversify, manually navigating the expanding design space becomes increasingly difficult. This paper aims to build an AI4AI framework for visual-token pruning by addressing a natural question: Can large language models automatically design effective visual-token reduction algorithms? Although LLMs possess broad algorithmic knowledge and strong reasoning capabilities, translating such general knowledge into effective solutions for a specialized task remains nontrivial. We argue that the key lies in designing an appropriate search-state representation that connects the internal knowledge of LLMs with the structural requirements and constraints of visual-token pruning. Based on this insight, we propose AutoPrune, a training-free framework for LLM-driven visual-token pruning policy design. At its core, AutoPrune introduces a Token Pruning Domain-Specific Language (TPDSL) comprising 131 reusable atoms for budget control, token scoring, selection constraints, and token reassembly. A key property of TPDSL is that it represents each search state as a residual modification of a strong base policy. This residual formulation narrows the search space and directs the LLM's attention toward the policy components that are most consequential for performance. Experiments on 14 multimodal benchmarks and three MLLM backbones demonstrate the effectiveness, efficiency, and transferability of AutoPrune. Even when removing 94.4% of visual tokens, AutoPrune preserves more than 99% of full-token performance while reducing FLOPs by 9.9x and prefill latency by 6.4x.","upvotes":6,"discussionId":"6a7eb64542823931a1f1779f","ai_summary":"AutoPrune uses large language models to automatically design visual-token pruning policies for multimodal models via a domain-specific language and residual search formulation, achieving high efficiency with minimal performance loss.","ai_keywords":["visual-token pruning","multimodal large language models","AutoPrune","TPDSL","token pruning domain-specific language","residual modification","base policy","FLOPs","prefill latency"],"ai_summary_model":"thinkingmachines/Inkling-Small"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"659cf9791c8b66637e3de72d","avatarUrl":"/avatars/7d26710f687be9444796980662614f16.svg","isPro":false,"fullname":"zhiqin yang","user":"visity","type":"user"},{"_id":"68ad5317ee6c21d50d189a9e","avatarUrl":"/avatars/f115c871a32065c431c72f4ad8295d2f.svg","isPro":false,"fullname":"Jingwen Fu","user":"Jamon123","type":"user"},{"_id":"64ecb174f22081b4ac7ca397","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/64ecb174f22081b4ac7ca397/PiAPtD_rbuhGOqfE6ZSIu.jpeg","isPro":false,"fullname":"Yunlong Lin","user":"LYL1015","type":"user"},{"_id":"646def60df618b303b419323","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/646def60df618b303b419323/JLJGYen4-5M8ivsLsSk0w.jpeg","isPro":false,"fullname":"Lei Wang","user":"demolei","type":"user"},{"_id":"69afe31f3984cb438460ab7f","avatarUrl":"/avatars/66de6966ba7662e8670a4b6debbc0acb.svg","isPro":false,"fullname":"Yuhan Liu","user":"Yuhan88","type":"user"},{"_id":"65905af887944e494e37e09a","avatarUrl":"/avatars/8f079ef18d06f506b94edca7d49e4c26.svg","isPro":false,"fullname":"Hert4","user":"beyoru","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.07193.md","query":{}}">

Get this paper in your agent:

Don't have the latest CLI?

Cite arxiv.org/abs/2608.07193 in a model README.md to link it from this page.

Cite arxiv.org/abs/2608.07193 in a dataset README.md to link it from this page.

Cite arxiv.org/abs/2608.07193 in a Space README.md to link it from this page.

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