{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OA7KLRV4HOEZ3NMINSQB2WBOSX","short_pith_number":"pith:OA7KLRV4","schema_version":"1.0","canonical_sha256":"703ea5c6bc3b899db5886ca01d582e95d47de7674dcd3215a243ba77cc8114f7","source":{"kind":"arxiv","id":"2603.19312","version":2},"attestation_state":"computed","paper":{"title":"LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels","license":"http://creativecommons.org/licenses/by/4.0/","headline":"LeWorldModel trains the first stable end-to-end JEPA from raw pixels using only two loss terms.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Damien Scieur, Lucas Maes, Quentin Le Lidec, Randall Balestriero, Yann LeCun","submitted_at":"2026-03-13T19:48:14Z","abstract_excerpt":"Joint Embedding Predictive Architectures (JEPAs) offer a compelling framework for learning world models in compact latent spaces, yet existing methods remain fragile, relying on complex multi-term losses, exponential moving averages, pre-trained encoders, or auxiliary supervision to avoid representation collapse. In this work, we introduce LeWorldModel (LeWM), the first JEPA that trains stably end-to-end from raw pixels using only two loss terms: a next-embedding prediction loss and a regularizer enforcing Gaussian-distributed latent embeddings. This reduces tunable loss hyperparameters from s"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2603.19312","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-13T19:48:14Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c996afe7bd20ae820ffebacb3d8545c079f16c9dbec5cd86e4964808eaa7e042","abstract_canon_sha256":"6a9f8b9150e3eed0763657a4c826c7ce16de4fc43189b9aea2b3d38f9c9389c8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:53.585343Z","signature_b64":"i6dVz7U7s1aHVxYy9CB96k8DS/Kj2hz9pEZmae8NSK/spMf2GLaFweu1LzWHU/kL2puPChTcDEOPcCbLCMZ1DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"703ea5c6bc3b899db5886ca01d582e95d47de7674dcd3215a243ba77cc8114f7","last_reissued_at":"2026-05-17T23:38:53.584673Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:53.584673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixels","license":"http://creativecommons.org/licenses/by/4.0/","headline":"LeWorldModel trains the first stable end-to-end JEPA from raw pixels using only two loss terms.","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Damien Scieur, Lucas Maes, Quentin Le Lidec, Randall Balestriero, Yann LeCun","submitted_at":"2026-03-13T19:48:14Z","abstract_excerpt":"Joint Embedding Predictive Architectures (JEPAs) offer a compelling framework for learning world models in compact latent spaces, yet existing methods remain fragile, relying on complex multi-term losses, exponential moving averages, pre-trained encoders, or auxiliary supervision to avoid representation collapse. In this work, we introduce LeWorldModel (LeWM), the first JEPA that trains stably end-to-end from raw pixels using only two loss terms: a next-embedding prediction loss and a regularizer enforcing Gaussian-distributed latent embeddings. This reduces tunable loss hyperparameters from s"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"LeWM is the first JEPA that trains stably end-to-end from raw pixels using only two loss terms: a next-embedding prediction loss and a regularizer enforcing Gaussian-distributed latent embeddings, reducing tunable loss hyperparameters from six to one.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the Gaussian regularizer alone is sufficient to prevent representation collapse across diverse 2D and 3D control tasks without any auxiliary supervision or pre-trained encoders.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LeWM is the first end-to-end trainable JEPA from pixels that uses only two loss terms for stable training and fast planning on 2D/3D control tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LeWorldModel trains the first stable end-to-end JEPA from raw pixels using only two loss terms.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e893dc285675daa050d70f68e802f6128c4f622eb21072acfc2ffbfad5d01c5a"},"source":{"id":"2603.19312","kind":"arxiv","version":2},"verdict":{"id":"0032a5ee-32be-4bf5-a7c6-5e61edf10d97","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T04:03:29.162452Z","strongest_claim":"LeWM is the first JEPA that trains stably end-to-end from raw pixels using only two loss terms: a next-embedding prediction loss and a regularizer enforcing Gaussian-distributed latent embeddings, reducing tunable loss hyperparameters from six to one.","one_line_summary":"LeWM is the first end-to-end trainable JEPA from pixels that uses only two loss terms for stable training and fast planning on 2D/3D control tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the Gaussian regularizer alone is sufficient to prevent representation collapse across diverse 2D and 3D control tasks without any auxiliary supervision or pre-trained encoders.","pith_extraction_headline":"LeWorldModel trains the first stable end-to-end JEPA from raw pixels using only two loss terms."},"references":{"count":57,"sample":[{"doi":"","year":2016,"title":"End-to-end training of deep visuomotor policies.Journal of Machine Learning Research, 17(39):1–40, 2016","work_id":"a556dd56-01e4-4727-a65b-696c8a91ddfb","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"World Models","work_id":"07227eee-8445-4c98-bce4-c6a6fd5ed907","ref_index":2,"cited_arxiv_id":"1803.10122","is_internal_anchor":true},{"doi":"","year":2023,"title":"Transformers are sample-efficient world models","work_id":"b39e8f37-9056-4866-a159-dacf3389c224","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Training Agents Inside of Scalable World Models","work_id":"f0464a07-aaee-486f-a0b1-a4bce0bbc3e4","ref_index":4,"cited_arxiv_id":"2509.24527","is_internal_anchor":true},{"doi":"","year":2022,"title":"A path towards autonomous machine intelligence version 0.9","work_id":"be31d909-e137-4d1d-ac0a-d54910a75946","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":57,"snapshot_sha256":"b7cdc6d899ac61f44d8afa22af899cfd96657609a030a6b451fadd754fcfd549","internal_anchors":13},"formal_canon":{"evidence_count":3,"snapshot_sha256":"478ca22640ab1adaea918b9964ea55e158037c87adcfd2bf41c5e988034a4937"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2603.19312","created_at":"2026-05-17T23:38:53.584787+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.19312v2","created_at":"2026-05-17T23:38:53.584787+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.19312","created_at":"2026-05-17T23:38:53.584787+00:00"},{"alias_kind":"pith_short_12","alias_value":"OA7KLRV4HOEZ","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"OA7KLRV4HOEZ3NMI","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"OA7KLRV4","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":28,"internal_anchor_count":28,"sample":[{"citing_arxiv_id":"2604.05157","citing_title":"IntentScore: Intent-Conditioned Action Evaluation for Computer-Use Agents","ref_index":12,"is_internal_anchor":true},{"citing_arxiv_id":"2605.23878","citing_title":"LaMo: Self-Supervised Latent Motion Priors for Physical Realism in Video Generation","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15153","citing_title":"Pelican-Unify 1.0: A Unified Embodied Intelligence Model for Understanding, Reasoning, Imagination and Action","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21800","citing_title":"stable-worldmodel: A Platform for Reproducible World Modeling Research and Evaluation","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22164","citing_title":"Beyond Euclidean Proximity: Repairing Latent World Models with Horizon-Matched Trajectory Reachability Metrics","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21963","citing_title":"ChronoMedicalWorld: A Medical World Model for Learning Patient Trajectories from Longitudinal Care Data","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16412","citing_title":"SCAR: Self-Supervised Continuous Action Representation Learning","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2605.15153","citing_title":"Pelican-Unify 1.0: A Unified Embodied Intelligence Model for Understanding, Reasoning, Imagination and Action","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13013","citing_title":"JEDI: Joint Embedding Diffusion World Model for Online Model-Based Reinforcement Learning","ref_index":23,"is_internal_anchor":true},{"citing_arxiv_id":"2603.28489","citing_title":"Video Generation Models as World Models: Efficient Paradigms, Architectures and Algorithms","ref_index":224,"is_internal_anchor":true},{"citing_arxiv_id":"2603.29496","citing_title":"Metriplector: From Field Theory to Neural Architecture","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.27411","citing_title":"Detecting is Easy, Adapting is Hard: Local Expert Growth for Visual Model-Based Reinforcement Learning under Distribution Shift","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09693","citing_title":"Do multimodal models imagine electric sheep?","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24662","citing_title":"Information bottleneck for learning the phase space of dynamics from high-dimensional experimental data","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2605.03413","citing_title":"Learning to Theorize the World from Observation","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06298","citing_title":"Render, Don't Decode: Weight-Space World Models with Latent Structural Disentanglement","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05586","citing_title":"AeroJEPA: Learning Semantic Latent Representations for Scalable 3D Aerodynamic Field Modeling","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.00080","citing_title":"World Model for Robot Learning: A Comprehensive Survey","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2604.11302","citing_title":"3D-Anchored Lookahead Planning for Persistent Robotic Scene Memory via World-Model-Based MCTS","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01694","citing_title":"Latent State Design for World Models under Sufficiency Constraints","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06841","citing_title":"AGWM: Affordance-Grounded World Models for Environments with Compositional Prerequisites","ref_index":64,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06298","citing_title":"Render, Don't Decode: Weight-Space World Models with Latent Structural Disentanglement","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07390","citing_title":"ST-Gen4D: Embedding 4D Spatiotemporal Cognition into World Model for 4D Generation","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07278","citing_title":"Predictive but Not Plannable: RC-aux for Latent World Models","ref_index":29,"is_internal_anchor":true},{"citing_arxiv_id":"2605.07514","citing_title":"Is the Future Compatible? Diagnosing Dynamic Consistency in World Action Models","ref_index":28,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":3,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX","json":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX.json","graph_json":"https://pith.science/api/pith-number/OA7KLRV4HOEZ3NMINSQB2WBOSX/graph.json","events_json":"https://pith.science/api/pith-number/OA7KLRV4HOEZ3NMINSQB2WBOSX/events.json","paper":"https://pith.science/paper/OA7KLRV4"},"agent_actions":{"view_html":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX","download_json":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX.json","view_paper":"https://pith.science/paper/OA7KLRV4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.19312&json=true","fetch_graph":"https://pith.science/api/pith-number/OA7KLRV4HOEZ3NMINSQB2WBOSX/graph.json","fetch_events":"https://pith.science/api/pith-number/OA7KLRV4HOEZ3NMINSQB2WBOSX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX/action/storage_attestation","attest_author":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX/action/author_attestation","sign_citation":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX/action/citation_signature","submit_replication":"https://pith.science/pith/OA7KLRV4HOEZ3NMINSQB2WBOSX/action/replication_record"}},"created_at":"2026-05-17T23:38:53.584787+00:00","updated_at":"2026-05-17T23:38:53.584787+00:00"}