{"paper":{"title":"RoboDreamer: Learning Compositional World Models for Robot Imagination","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"RoboDreamer factorizes video generation using language primitives to create plans for unseen robot tasks.","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Chuang Gan, Dit-Yan Yeung, Jiaben Chen, Siyuan Zhou, Yandong Li, Yilun Du","submitted_at":"2024-04-18T17:58:03Z","abstract_excerpt":"Text-to-video models have demonstrated substantial potential in robotic decision-making, enabling the imagination of realistic plans of future actions as well as accurate environment simulation. However, one major issue in such models is generalization -- models are limited to synthesizing videos subject to language instructions similar to those seen at training time. This is heavily limiting in decision-making, where we seek a powerful world model to synthesize plans of unseen combinations of objects and actions in order to solve previously unseen tasks in new environments. To resolve this is"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our approach can successfully synthesize video plans on unseen goals in the RT-X, enables successful robot execution in simulation, and substantially outperforms monolithic baseline approaches to video generation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That natural language instructions can be reliably parsed into lower-level primitives whose separate models compose into coherent, realistic videos without introducing artifacts or losing task-relevant details.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RoboDreamer factorizes video generation using language primitives to achieve compositional generalization in robot world models, outperforming monolithic baselines on unseen goals in RT-X.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"RoboDreamer factorizes video generation using language primitives to create plans for unseen robot tasks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"98546809f5544792a6442e94ac62eae53b46ea6b6b1221ea558c9a5d60fcd642"},"source":{"id":"2404.12377","kind":"arxiv","version":1},"verdict":{"id":"cd704efb-c6fd-468c-992c-3d2d32701d49","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T20:43:49.453064Z","strongest_claim":"Our approach can successfully synthesize video plans on unseen goals in the RT-X, enables successful robot execution in simulation, and substantially outperforms monolithic baseline approaches to video generation.","one_line_summary":"RoboDreamer factorizes video generation using language primitives to achieve compositional generalization in robot world models, outperforming monolithic baselines on unseen goals in RT-X.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That natural language instructions can be reliably parsed into lower-level primitives whose separate models compose into coherent, realistic videos without introducing artifacts or losing task-relevant details.","pith_extraction_headline":"RoboDreamer factorizes video generation using language primitives to create plans for unseen robot tasks."},"references":{"count":65,"sample":[{"doi":"","year":2021,"title":"Unsupervised learning of compositional energy concepts","work_id":"585f2b86-ddff-414d-ada6-8a36e14fef70","ref_index":7,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"B., Dieleman, S., Fergus, R., Sohl-Dickstein, J., Doucet, A., and Grathwohl, W","work_id":"1bf90112-a203-409f-8900-11651a71919b","ref_index":8,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"G., Tapaswi, M., Laptev, I., and Schmid, C","work_id":"646c92f1-d546-4fbb-8968-71d5ee6f9718","ref_index":12,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Diffusion-based generation, optimization, and planning in 3d scenes","work_id":"6bff28de-131c-49cc-86fa-2b61c4212c98","ref_index":15,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"R., and Davison, A","work_id":"ebc7a356-db9d-48bd-8ae8-be7274070335","ref_index":17,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":65,"snapshot_sha256":"d9f370a51d0cca10de936e6bd758f85f4a968cc6ab1eb5a137c4add3f2f0e6f7","internal_anchors":12},"formal_canon":{"evidence_count":2,"snapshot_sha256":"c256a1870659fa9e3d01b953c60dd7a4b6a37d8529245a7b3f095707f4136afd"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}