Hugging Face Daily Papers · · 4 min read

VLMs are Good Teachers for Video Reasoning via Adaptive Test-Time Optimization

Mirrored from Hugging Face Daily Papers for archival readability. Support the source by reading on the original site.

Video Generation Models can produce temporally coherent visual trajectories, yet often fail to follow task-specific rules. We introduce a VLM-as-Teacher framework that synthesizes task-specific reward queries and guides a VGM Reasoner through online test-time optimization of a lightweight LoRA module.</p>\n","updatedAt":"2026-06-02T03:31:58.635Z","author":{"_id":"6506b77a773ceaa8d52ecea1","avatarUrl":"/avatars/0e769a0795063e1491c44760a4a83097.svg","fullname":"CJH","name":"Howe666","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.7915430068969727},"editors":["Howe666"],"editorAvatarUrls":["/avatars/0e769a0795063e1491c44760a4a83097.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.02564","authors":[{"_id":"6a1e4e7a808ddbc3c7d43d19","name":"Junhao Cheng","hidden":false},{"_id":"6a1e4e7a808ddbc3c7d43d1a","name":"Liang Hou","hidden":false},{"_id":"6a1e4e7a808ddbc3c7d43d1b","name":"Tianxiong Zhong","hidden":false},{"_id":"6a1e4e7a808ddbc3c7d43d1c","name":"Xin Tao","hidden":false},{"_id":"6a1e4e7a808ddbc3c7d43d1d","name":"Pengfei Wan","hidden":false},{"_id":"6a1e4e7a808ddbc3c7d43d1e","name":"Kun Gai","hidden":false},{"_id":"6a1e4e7a808ddbc3c7d43d1f","name":"Jing Liao","hidden":false}],"publishedAt":"2026-06-01T00:00:00.000Z","submittedOnDailyAt":"2026-06-02T00:00:00.000Z","title":"VLMs are Good Teachers for Video Reasoning via Adaptive Test-Time Optimization","submittedOnDailyBy":{"_id":"6506b77a773ceaa8d52ecea1","avatarUrl":"/avatars/0e769a0795063e1491c44760a4a83097.svg","isPro":false,"fullname":"CJH","user":"Howe666","type":"user","name":"Howe666"},"summary":"The recent \"Reasoning with Video\" paradigm utilizes Video Generation Models (VGMs) to generate temporally coherent visual trajectories to complete reasoning tasks. Although state-of-the-art VGMs excel at visual quality, they often struggle to understand and follow task-specific rules, leading to logical failures across diverse reasoning scenarios. Existing efforts try to utilize Vision-Language Models (VLMs) as problem pre-solvers to produce or refine textual guidance for the VGM. However, textual descriptions fail to capture intricate spatiotemporal details, and VGMs often struggle to faithfully execute fine-grained or long-tail instructions even with a valid plan. While VLMs struggle as solvers, they possess strong perception capabilities to evaluate process-constraint satisfaction and final-goal achievement. Leveraging this strength, we introduce a paradigm shift that transitions the role of VLMs to \"teachers\". Specifically, a VLM teacher extracts task-specific rules to formulate differentiable rewards, guiding a VGM Reasoner via test-time online optimization of a lightweight LoRA module. This strategy enables adaptive test-time optimization and extends the reasoning capabilities beyond the VGM's intrinsic boundaries. Evaluations on symbolic (VBVR-Bench) and general-purpose (RULER-Bench) video reasoning benchmarks show that the proposed method yields a 16.7-point average performance gain, outperforming the VLM-as-Solver paradigm (+0.4 points) and Best-of-N scaling (+2.2 points) by a large margin at comparable test-time cost. These findings reveal that integrating VLMs as test-time teachers offers a promising paradigm for achieving generalizable video reasoning. Project Page: https://VLM-as-Teacher.github.io/","upvotes":19,"discussionId":"6a1e4e7a808ddbc3c7d43d20","projectPage":"https://vlm-as-teacher.github.io/","ai_summary":"Video generation models combined with vision-language models acting as test-time teachers through differentiable rewards achieve superior video reasoning performance.","ai_keywords":["Video Generation Models","Vision-Language Models","test-time optimization","LoRA module","differentiable rewards","video reasoning benchmarks","VBVR-Bench","RULER-Bench"],"organization":{"_id":"662c559b322afcbae51b3c8b","name":"KlingTeam","fullname":"Kling Team","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e272ca6c78a8c122b12127/ZQV1aKLUDPf2rUcxxAqj6.jpeg"}},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6506b77a773ceaa8d52ecea1","avatarUrl":"/avatars/0e769a0795063e1491c44760a4a83097.svg","isPro":false,"fullname":"CJH","user":"Howe666","type":"user"},{"_id":"67ed5a9edf57c2d2b70f642d","avatarUrl":"/avatars/6483ca52c3d9fcd9c99917f9b5ce64f7.svg","isPro":false,"fullname":"Jiageng Wang","user":"TheWind-upBird","type":"user"},{"_id":"6826a05566f27e294e84507c","avatarUrl":"/avatars/ed611fab42227bab6a7584ecfdfa6f59.svg","isPro":false,"fullname":"abc","user":"pqowieuryrty","type":"user"},{"_id":"668a17d58c440fe1956448ba","avatarUrl":"/avatars/16fdfcdba1a7819f95c55c031ab27de5.svg","isPro":false,"fullname":"Xi Lu","user":"ln-e1","type":"user"},{"_id":"620783f24e28382272337ba4","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/620783f24e28382272337ba4/zkUveQPNiDfYjgGhuFErj.jpeg","isPro":false,"fullname":"GuoLiangTang","user":"Tommy930","type":"user"},{"_id":"662f93942510ef5735d7ad00","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/662f93942510ef5735d7ad00/ZIDIPm63sncIHFTT5b0uR.png","isPro":false,"fullname":"magicwpf","user":"magicwpf","type":"user"},{"_id":"6039478ab3ecf716b1a5fd4d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6039478ab3ecf716b1a5fd4d/_Thy4E7taiSYBLKxEKJbT.jpeg","isPro":true,"fullname":"taesiri","user":"taesiri","type":"user"},{"_id":"69cceed2eb9cdf88f2a50b65","avatarUrl":"/avatars/07cb3a54017fa3cfd87e0e6a65969e6f.svg","isPro":false,"fullname":"Qianyi Tang","user":"jacobrodriguezu","type":"user"},{"_id":"69ce390201d713064aea5864","avatarUrl":"/avatars/af3977aeb5432599fb6b576c3f64a46b.svg","isPro":false,"fullname":"Bohan Zeng","user":"zbhpku1","type":"user"},{"_id":"668df98de9e585e8718f767f","avatarUrl":"/avatars/2be52f4ae88a0991c8ae584f8e870734.svg","isPro":false,"fullname":"Xiangyang Luo","user":"XiangyangLuo02","type":"user"},{"_id":"69213991ce9fdaab0fec99e6","avatarUrl":"/avatars/9261b298aeee2c655747eb7c9a889754.svg","isPro":false,"fullname":"mini","user":"ge4edu","type":"user"},{"_id":"69cd3de429da70aa42016d49","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/wMKvVTPKlkKqFmChfDYif.png","isPro":false,"fullname":"Xinyu Zhang","user":"gaohaoyu","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"organization":{"_id":"662c559b322afcbae51b3c8b","name":"KlingTeam","fullname":"Kling Team","avatar":"https://cdn-avatars.huggingface.co/v1/production/uploads/60e272ca6c78a8c122b12127/ZQV1aKLUDPf2rUcxxAqj6.jpeg"},"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.02564.md"}">
Papers
arxiv:2606.02564

VLMs are Good Teachers for Video Reasoning via Adaptive Test-Time Optimization

Published on Jun 1
· Submitted by
CJH
on Jun 2
Authors:
,
,
,
,
,
,

Abstract

Video generation models combined with vision-language models acting as test-time teachers through differentiable rewards achieve superior video reasoning performance.

AI-generated summary

The recent "Reasoning with Video" paradigm utilizes Video Generation Models (VGMs) to generate temporally coherent visual trajectories to complete reasoning tasks. Although state-of-the-art VGMs excel at visual quality, they often struggle to understand and follow task-specific rules, leading to logical failures across diverse reasoning scenarios. Existing efforts try to utilize Vision-Language Models (VLMs) as problem pre-solvers to produce or refine textual guidance for the VGM. However, textual descriptions fail to capture intricate spatiotemporal details, and VGMs often struggle to faithfully execute fine-grained or long-tail instructions even with a valid plan. While VLMs struggle as solvers, they possess strong perception capabilities to evaluate process-constraint satisfaction and final-goal achievement. Leveraging this strength, we introduce a paradigm shift that transitions the role of VLMs to "teachers". Specifically, a VLM teacher extracts task-specific rules to formulate differentiable rewards, guiding a VGM Reasoner via test-time online optimization of a lightweight LoRA module. This strategy enables adaptive test-time optimization and extends the reasoning capabilities beyond the VGM's intrinsic boundaries. Evaluations on symbolic (VBVR-Bench) and general-purpose (RULER-Bench) video reasoning benchmarks show that the proposed method yields a 16.7-point average performance gain, outperforming the VLM-as-Solver paradigm (+0.4 points) and Best-of-N scaling (+2.2 points) by a large margin at comparable test-time cost. These findings reveal that integrating VLMs as test-time teachers offers a promising paradigm for achieving generalizable video reasoning. Project Page: https://VLM-as-Teacher.github.io/

Community

Paper submitter about 6 hours ago

Video Generation Models can produce temporally coherent visual trajectories, yet often fail to follow task-specific rules. We introduce a VLM-as-Teacher framework that synthesizes task-specific reward queries and guides a VGM Reasoner through online test-time optimization of a lightweight LoRA module.

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images

· Sign up or log in to comment

Get this paper in your agent:

hf papers read 2606.02564
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2606.02564 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2606.02564 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2606.02564 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection 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.

More from Hugging Face Daily Papers