{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:E25GRCMEVIVYKWIEQV37DKGIBP","short_pith_number":"pith:E25GRCME","canonical_record":{"source":{"id":"2605.01896","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-03T14:22:22Z","cross_cats_sorted":[],"title_canon_sha256":"c5197caf6bb6e357e1fde407eede488534ed061bd7b76c85ade173ceefd6409d","abstract_canon_sha256":"b7181e8bcd07ac9db4ee641caa78ccfccb5c06f9ea71f102d4f9c609e077a1fc"},"schema_version":"1.0"},"canonical_sha256":"26ba688984aa2b8559048577f1a8c80bf6317aec6c146228e1bf54ec6a9a7857","source":{"kind":"arxiv","id":"2605.01896","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01896","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01896v2","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01896","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"pith_short_12","alias_value":"E25GRCMEVIVY","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"pith_short_16","alias_value":"E25GRCMEVIVYKWIE","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"pith_short_8","alias_value":"E25GRCME","created_at":"2026-07-02T01:17:32Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:E25GRCMEVIVYKWIEQV37DKGIBP","target":"record","payload":{"canonical_record":{"source":{"id":"2605.01896","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-03T14:22:22Z","cross_cats_sorted":[],"title_canon_sha256":"c5197caf6bb6e357e1fde407eede488534ed061bd7b76c85ade173ceefd6409d","abstract_canon_sha256":"b7181e8bcd07ac9db4ee641caa78ccfccb5c06f9ea71f102d4f9c609e077a1fc"},"schema_version":"1.0"},"canonical_sha256":"26ba688984aa2b8559048577f1a8c80bf6317aec6c146228e1bf54ec6a9a7857","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-02T01:17:32.038218Z","signature_b64":"Sf7r5sPaeyWi9BJcFFoUzc+8kWI1ft9V7yLLoDIHOwi9bvCMj5w+BmE8jOWgWuyh+nE/+cz4T857/pe632OgBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"26ba688984aa2b8559048577f1a8c80bf6317aec6c146228e1bf54ec6a9a7857","last_reissued_at":"2026-07-02T01:17:32.037792Z","signature_status":"signed_v1","first_computed_at":"2026-07-02T01:17:32.037792Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.01896","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-02T01:17:32Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YNoKY6BCsRA+1eHxRdyzPwd/rh/hZFG2AWsj9RTl5vrhZ7vUBacObOPo0najq8xEjGiktf5uW4rS8c7XR6NlCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T15:56:07.874228Z"},"content_sha256":"77ba6c3b03663c6c35fa3a75f06ef85c80de434286591d8639b847b9b432d8ca","schema_version":"1.0","event_id":"sha256:77ba6c3b03663c6c35fa3a75f06ef85c80de434286591d8639b847b9b432d8ca"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:E25GRCMEVIVYKWIEQV37DKGIBP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Divide and Conquer: Decoupled Representation Alignment for Multimodal World Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Decoupling modality features in a diffusion model allows each to align with a separate foundation model expert for improved multi-modal video generation.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bin Xia, Dingkang Liang, Guangmo Yi, Jianlou Si, Jun Huang, Junyuan Xiao, Qiang Lyu, Shurui Shi, Tongtong Su, Wenming Yang, Xin Zhou, Yixuan Ye","submitted_at":"2026-05-03T14:22:22Z","abstract_excerpt":"Emerging multi-modal world models attempt to jointly generate videos across diverse modalities (e.g., RGB, depth, and mask), yet they fail to fully exploit the rich priors of existing foundation models. We propose $M^2$-REPA, the first representation alignment method tailored for multi-modal video generation. Our key insight is that foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary \"experts.\" Specifically, we first decouple modality-specific features from the diffusion model's intermediate representations, then ali"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"This design enables joint optimization, fully exploiting priors from multiple foundation models. Extensive experiments demonstrate that our method significantly outperforms baselines in visual quality and long-term consistency.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"Foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary experts, and that the proposed decoupling regularization will maintain this complementarity without introducing new conflicts during joint optimization.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"M²-REPA decouples modality-specific features inside a diffusion model and aligns each to its matching expert foundation model via an alignment loss plus a decoupling regularizer, yielding better visual quality and long-term consistency in multi-modal video generation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Decoupling modality features in a diffusion model allows each to align with a separate foundation model expert for improved multi-modal video generation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9efa3b9952af6078f2dbaf49bbe5568a80b203fe15b40d87115281738e8fca9a"},"source":{"id":"2605.01896","kind":"arxiv","version":2},"verdict":{"id":"2f394438-cabb-4762-8192-90acf0ea93bf","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T15:47:58.524895Z","strongest_claim":"This design enables joint optimization, fully exploiting priors from multiple foundation models. Extensive experiments demonstrate that our method significantly outperforms baselines in visual quality and long-term consistency.","one_line_summary":"M²-REPA decouples modality-specific features inside a diffusion model and aligns each to its matching expert foundation model via an alignment loss plus a decoupling regularizer, yielding better visual quality and long-term consistency in multi-modal video generation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"Foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary experts, and that the proposed decoupling regularization will maintain this complementarity without introducing new conflicts during joint optimization.","pith_extraction_headline":"Decoupling modality features in a diffusion model allows each to align with a separate foundation model expert for improved multi-modal video generation."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.01896/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T17:33:29.025209Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T04:31:23.000503Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T16:55:12.035885Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"9c61b9c96d00f86de8f33717c3ed61ec85d493659baf66c37f60d129e92021ba"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"2f394438-cabb-4762-8192-90acf0ea93bf"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-02T01:17:32Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"cMsy25AZZyUCvXV9NWjkYJUgm3P1zNkxRURmgP9IDWeVx1RA+U77kBDyyntmNe+IXKn97F+eWO9eL9QWYbmVCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T15:56:07.876908Z"},"content_sha256":"7bd6f6798989fda9bb71f841805404d480d75c5cd4addd8e2bdf9cecf04126b3","schema_version":"1.0","event_id":"sha256:7bd6f6798989fda9bb71f841805404d480d75c5cd4addd8e2bdf9cecf04126b3"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/E25GRCMEVIVYKWIEQV37DKGIBP/bundle.json","state_url":"https://pith.science/pith/E25GRCMEVIVYKWIEQV37DKGIBP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/E25GRCMEVIVYKWIEQV37DKGIBP/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-03T15:56:07Z","links":{"resolver":"https://pith.science/pith/E25GRCMEVIVYKWIEQV37DKGIBP","bundle":"https://pith.science/pith/E25GRCMEVIVYKWIEQV37DKGIBP/bundle.json","state":"https://pith.science/pith/E25GRCMEVIVYKWIEQV37DKGIBP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/E25GRCMEVIVYKWIEQV37DKGIBP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:E25GRCMEVIVYKWIEQV37DKGIBP","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"b7181e8bcd07ac9db4ee641caa78ccfccb5c06f9ea71f102d4f9c609e077a1fc","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-03T14:22:22Z","title_canon_sha256":"c5197caf6bb6e357e1fde407eede488534ed061bd7b76c85ade173ceefd6409d"},"schema_version":"1.0","source":{"id":"2605.01896","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01896","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01896v2","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01896","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"pith_short_12","alias_value":"E25GRCMEVIVY","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"pith_short_16","alias_value":"E25GRCMEVIVYKWIE","created_at":"2026-07-02T01:17:32Z"},{"alias_kind":"pith_short_8","alias_value":"E25GRCME","created_at":"2026-07-02T01:17:32Z"}],"graph_snapshots":[{"event_id":"sha256:7bd6f6798989fda9bb71f841805404d480d75c5cd4addd8e2bdf9cecf04126b3","target":"graph","created_at":"2026-07-02T01:17:32Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"This design enables joint optimization, fully exploiting priors from multiple foundation models. Extensive experiments demonstrate that our method significantly outperforms baselines in visual quality and long-term consistency."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"Foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary experts, and that the proposed decoupling regularization will maintain this complementarity without introducing new conflicts during joint optimization."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"M²-REPA decouples modality-specific features inside a diffusion model and aligns each to its matching expert foundation model via an alignment loss plus a decoupling regularizer, yielding better visual quality and long-term consistency in multi-modal video generation."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Decoupling modality features in a diffusion model allows each to align with a separate foundation model expert for improved multi-modal video generation."}],"snapshot_sha256":"9efa3b9952af6078f2dbaf49bbe5568a80b203fe15b40d87115281738e8fca9a"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T17:33:29.025209Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-20T04:31:23.000503Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T16:55:12.035885Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.01896/integrity.json","findings":[],"snapshot_sha256":"9c61b9c96d00f86de8f33717c3ed61ec85d493659baf66c37f60d129e92021ba","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Emerging multi-modal world models attempt to jointly generate videos across diverse modalities (e.g., RGB, depth, and mask), yet they fail to fully exploit the rich priors of existing foundation models. We propose $M^2$-REPA, the first representation alignment method tailored for multi-modal video generation. Our key insight is that foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary \"experts.\" Specifically, we first decouple modality-specific features from the diffusion model's intermediate representations, then ali","authors_text":"Bin Xia, Dingkang Liang, Guangmo Yi, Jianlou Si, Jun Huang, Junyuan Xiao, Qiang Lyu, Shurui Shi, Tongtong Su, Wenming Yang, Xin Zhou, Yixuan Ye","cross_cats":[],"headline":"Decoupling modality features in a diffusion model allows each to align with a separate foundation model expert for improved multi-modal video generation.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-03T14:22:22Z","title":"Divide and Conquer: Decoupled Representation Alignment for Multimodal World Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.01896","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T15:47:58.524895Z","id":"2f394438-cabb-4762-8192-90acf0ea93bf","model_set":{"reader":"grok-4.3"},"one_line_summary":"M²-REPA decouples modality-specific features inside a diffusion model and aligns each to its matching expert foundation model via an alignment loss plus a decoupling regularizer, yielding better visual quality and long-term consistency in multi-modal video generation.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Decoupling modality features in a diffusion model allows each to align with a separate foundation model expert for improved multi-modal video generation.","strongest_claim":"This design enables joint optimization, fully exploiting priors from multiple foundation models. Extensive experiments demonstrate that our method significantly outperforms baselines in visual quality and long-term consistency.","weakest_assumption":"Foundation models trained on different modality spaces naturally capture distinct domain-specific priors, acting as complementary experts, and that the proposed decoupling regularization will maintain this complementarity without introducing new conflicts during joint optimization."}},"verdict_id":"2f394438-cabb-4762-8192-90acf0ea93bf"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:77ba6c3b03663c6c35fa3a75f06ef85c80de434286591d8639b847b9b432d8ca","target":"record","created_at":"2026-07-02T01:17:32Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"b7181e8bcd07ac9db4ee641caa78ccfccb5c06f9ea71f102d4f9c609e077a1fc","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-03T14:22:22Z","title_canon_sha256":"c5197caf6bb6e357e1fde407eede488534ed061bd7b76c85ade173ceefd6409d"},"schema_version":"1.0","source":{"id":"2605.01896","kind":"arxiv","version":2}},"canonical_sha256":"26ba688984aa2b8559048577f1a8c80bf6317aec6c146228e1bf54ec6a9a7857","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"26ba688984aa2b8559048577f1a8c80bf6317aec6c146228e1bf54ec6a9a7857","first_computed_at":"2026-07-02T01:17:32.037792Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-02T01:17:32.037792Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Sf7r5sPaeyWi9BJcFFoUzc+8kWI1ft9V7yLLoDIHOwi9bvCMj5w+BmE8jOWgWuyh+nE/+cz4T857/pe632OgBw==","signature_status":"signed_v1","signed_at":"2026-07-02T01:17:32.038218Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.01896","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:77ba6c3b03663c6c35fa3a75f06ef85c80de434286591d8639b847b9b432d8ca","sha256:7bd6f6798989fda9bb71f841805404d480d75c5cd4addd8e2bdf9cecf04126b3"],"state_sha256":"fc0314e730ab33ca09f5fe8c35807f9ce15f3aa763378a904e9d3ce581d5d0b3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"05DaH58B8P35Upo+5lz5lFfGv2EghmA+O3iAE72jDYKAhoMnD4DU71XQSqMgOnoj1ZBr7HsRP0ENqMo7XZi5DA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T15:56:07.885512Z","bundle_sha256":"1818ba97da71f3923568c5a35ae1d101429da370164b8d2b5b6901bf1ac249e6"}}