Hugging Face Daily Papers · June 8, 2026 · 3 min read

Robots Need More than VLA and World Models

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

Like Read original ↗

VLAs are great, World Models are great. But Robots need much more!</p>\n","updatedAt":"2026-06-08T11:05:09.781Z","author":{"_id":"631c375768f7da9ad2496bf6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/631c375768f7da9ad2496bf6/1sDOoecA6e1v_hn_VAgUq.jpeg","fullname":"Haitham Bou Ammar","name":"hba123","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":97,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.9439261555671692},"editors":["hba123"],"editorAvatarUrls":["https://cdn-avatars.huggingface.co/v1/production/uploads/631c375768f7da9ad2496bf6/1sDOoecA6e1v_hn_VAgUq.jpeg"],"reactions":[{"reaction":"👍","users":["RasT009","hba123","MotoniqSteven","singhanshuman"],"count":4}],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2606.06556","authors":[{"_id":"6a26a1bfe4c258a02949231e","name":"Elis Karcini","hidden":false},{"_id":"6a26a1bfe4c258a02949231f","name":"Faisal Mehrban","hidden":false},{"_id":"6a26a1bfe4c258a029492320","name":"Quang Nguyen","hidden":false},{"_id":"6a26a1bfe4c258a029492321","name":"Mac Schwager","hidden":false},{"_id":"6a26a1bfe4c258a029492322","name":"Arash Ajoudani","hidden":false},{"_id":"6a26a1bfe4c258a029492323","name":"Cesar Cadena","hidden":false},{"_id":"6a26a1bfe4c258a029492324","name":"Jan Peters","hidden":false},{"_id":"6a26a1bfe4c258a029492325","name":"Marco Hutter","hidden":false},{"_id":"6a26a1bfe4c258a029492326","name":"Haitham Bou-Ammar","hidden":false}],"publishedAt":"2026-06-04T00:00:00.000Z","submittedOnDailyAt":"2026-06-08T00:00:00.000Z","title":"Robots Need More than VLA and World Models","submittedOnDailyBy":{"_id":"631c375768f7da9ad2496bf6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/631c375768f7da9ad2496bf6/1sDOoecA6e1v_hn_VAgUq.jpeg","isPro":true,"fullname":"Haitham Bou Ammar","user":"hba123","type":"user","name":"hba123"},"summary":"Generalist robot intelligence is often framed as a policy-scaling problem: collect more robot demonstrations, train larger Vision-Language-Action (VLA) models, and expect broader generalisation. In this position paper, we argue that this framing is incomplete. The central bottleneck is not only policy learning, but the absence of mechanisms that convert the world's abundant unstructured behavioural data into grounded robot supervision. Human motion, internet video, simulation rollouts, and interactive demonstrations contain rich information about tasks, goals, contacts, failures, and physical constraints, yet most of this information is not directly usable by robot policies because it lacks embodiment-specific action labels, task semantics, and reward structure. We identify four missing components for the next generation of robotics: data interfaces for autolabelling unstructured behaviour, embodiment interfaces for retargeting human motion to robot actions, world-model interfaces for physics-grounded 3D reasoning, and reward interfaces for inferring task progress and success from video and language. We survey recent progress in robot foundation models, cross-embodiment datasets, learning from video, world models, and reward modelling, and propose a research agenda for building robotics systems that can learn not only from robot demonstrations, but from the broader physical world.","upvotes":19,"discussionId":"6a26a1c0e4c258a029492327","ai_summary":"Robot intelligence advancement requires integrating unstructured behavioral data through specialized interfaces for labeling, embodiment mapping, world modeling, and reward inference rather than relying solely on policy scaling.","ai_keywords":["Vision-Language-Action models","robot demonstrations","embodied intelligence","cross-embodiment datasets","world models","reward modeling","autolabeling","retargeting","3D reasoning"],"ai_summary_model":"Qwen/Qwen2.5-Coder-32B-Instruct"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"631c375768f7da9ad2496bf6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/631c375768f7da9ad2496bf6/1sDOoecA6e1v_hn_VAgUq.jpeg","isPro":true,"fullname":"Haitham Bou Ammar","user":"hba123","type":"user"},{"_id":"61406a3a7ca990b6ad188a5d","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/61406a3a7ca990b6ad188a5d/oIILoSMhLuAjVMSY14vKj.jpeg","isPro":false,"fullname":"Ahmed Khaled Khamis","user":"KickItLikeShika","type":"user"},{"_id":"6a20357efea3f60347fb71ff","avatarUrl":"/avatars/f04fc9da083f36220334c3cb16ef75ae.svg","isPro":false,"fullname":"Gregory Kell","user":"greg-huawei","type":"user"},{"_id":"671880ff5f1f9948bf2ce2d1","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/qOQW6D05xIM1HPc-S_adk.png","isPro":false,"fullname":"Zafeirios Fountas","user":"zfountas","type":"user"},{"_id":"663940252e17a8ac963e20ca","avatarUrl":"/avatars/07d7bf27931552b8a577eff09b81f6bd.svg","isPro":false,"fullname":"S N","user":"Refinath","type":"user"},{"_id":"668407961cfe79e7fd5c885c","avatarUrl":"/avatars/2e68cd8379d57842ac5e21237c8ca70c.svg","isPro":false,"fullname":"Ras","user":"RasT009","type":"user"},{"_id":"67b71bc5622b213fdf9fe090","avatarUrl":"/avatars/9c986d725608b42596e71fb413236437.svg","isPro":false,"fullname":"Martin Benfeghoul","user":"MrTin1","type":"user"},{"_id":"667b6e28a0e1bc140664e725","avatarUrl":"/avatars/2e4a69a0f98ce6de3f044b2831609ea4.svg","isPro":false,"fullname":"xiaotong","user":"xtongji","type":"user"},{"_id":"6270324ebecab9e2dcf245de","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6270324ebecab9e2dcf245de/cMbtWSasyNlYc9hvsEEzt.jpeg","isPro":false,"fullname":"Kye Gomez","user":"kye","type":"user"},{"_id":"6a26b18e3ca36c4f8efc09b7","avatarUrl":"/avatars/0b917b91ad000597a110ff5fcb331cb9.svg","isPro":false,"fullname":"Olivier Ayache","user":"orayache","type":"user"},{"_id":"66c5beb29fc647b08321cc20","avatarUrl":"/avatars/d84e497cbcd44e7e06449092033f5bbd.svg","isPro":false,"fullname":"Matthieu Zimmer","user":"MatthieuZ","type":"user"},{"_id":"6a26c755a89de37b041fd791","avatarUrl":"/avatars/cfe352e794359819c0ee5850693b84a3.svg","isPro":false,"fullname":"Max R","user":"rossomax","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2606/2606.06556.md"}">

Papers

arxiv:2606.06556

Robots Need More than VLA and World Models

Published on Jun 4

· Submitted by

Haitham Bou Ammar on Jun 8

Upvote

Authors:

Abstract

Robot intelligence advancement requires integrating unstructured behavioral data through specialized interfaces for labeling, embodiment mapping, world modeling, and reward inference rather than relying solely on policy scaling.

Generated by Qwen/Qwen2.5-Coder-32B-Instruct

Generalist robot intelligence is often framed as a policy-scaling problem: collect more robot demonstrations, train larger Vision-Language-Action (VLA) models, and expect broader generalisation. In this position paper, we argue that this framing is incomplete. The central bottleneck is not only policy learning, but the absence of mechanisms that convert the world's abundant unstructured behavioural data into grounded robot supervision. Human motion, internet video, simulation rollouts, and interactive demonstrations contain rich information about tasks, goals, contacts, failures, and physical constraints, yet most of this information is not directly usable by robot policies because it lacks embodiment-specific action labels, task semantics, and reward structure. We identify four missing components for the next generation of robotics: data interfaces for autolabelling unstructured behaviour, embodiment interfaces for retargeting human motion to robot actions, world-model interfaces for physics-grounded 3D reasoning, and reward interfaces for inferring task progress and success from video and language. We survey recent progress in robot foundation models, cross-embodiment datasets, learning from video, world models, and reward modelling, and propose a research agenda for building robotics systems that can learn not only from robot demonstrations, but from the broader physical world.

View arXiv page View PDF Add to collection