{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:RSGS7RD25ZYA5DB3KSOFL32NJF","short_pith_number":"pith:RSGS7RD2","schema_version":"1.0","canonical_sha256":"8c8d2fc47aee700e8c3b549c55ef4d4960e5a97a7ea10cc3703758261d2383fe","source":{"kind":"arxiv","id":"2110.04121","version":2},"attestation_state":"computed","paper":{"title":"On the Limitations of Multimodal VAEs","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Emanuele Palumbo, Imant Daunhawer, Julia E. Vogt, Kieran Chin-Cheong, Thomas M. Sutter","submitted_at":"2021-10-08T13:28:28Z","abstract_excerpt":"Multimodal variational autoencoders (VAEs) have shown promise as efficient generative models for weakly-supervised data. Yet, despite their advantage of weak supervision, they exhibit a gap in generative quality compared to unimodal VAEs, which are completely unsupervised. In an attempt to explain this gap, we uncover a fundamental limitation that applies to a large family of mixture-based multimodal VAEs. We prove that the sub-sampling of modalities enforces an undesirable upper bound on the multimodal ELBO and thereby limits the generative quality of the respective models. Empirically, we sh"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2110.04121","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2021-10-08T13:28:28Z","cross_cats_sorted":[],"title_canon_sha256":"6074066a81d90840abebf3ec5060fc2a3ef8680a23558138e96d222186ba4ba5","abstract_canon_sha256":"c1b0a142a428a7b4e77a3a176b4ea46d84e3f4608107e4da50175d3b1928b42c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:12:28.443469Z","signature_b64":"GQeNzZqpwSMVySNdTB4n1QiRYzX9aMZ2355PORAJWyCLd8+dLaqaLapzq2WYUjLtqfFzS7KzSw9ruUTaZCQNCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8c8d2fc47aee700e8c3b549c55ef4d4960e5a97a7ea10cc3703758261d2383fe","last_reissued_at":"2026-07-05T04:12:28.443015Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:12:28.443015Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the Limitations of Multimodal VAEs","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Emanuele Palumbo, Imant Daunhawer, Julia E. Vogt, Kieran Chin-Cheong, Thomas M. Sutter","submitted_at":"2021-10-08T13:28:28Z","abstract_excerpt":"Multimodal variational autoencoders (VAEs) have shown promise as efficient generative models for weakly-supervised data. Yet, despite their advantage of weak supervision, they exhibit a gap in generative quality compared to unimodal VAEs, which are completely unsupervised. In an attempt to explain this gap, we uncover a fundamental limitation that applies to a large family of mixture-based multimodal VAEs. We prove that the sub-sampling of modalities enforces an undesirable upper bound on the multimodal ELBO and thereby limits the generative quality of the respective models. Empirically, we sh"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.04121","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2110.04121/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2110.04121","created_at":"2026-07-05T04:12:28.443072+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.04121v2","created_at":"2026-07-05T04:12:28.443072+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.04121","created_at":"2026-07-05T04:12:28.443072+00:00"},{"alias_kind":"pith_short_12","alias_value":"RSGS7RD25ZYA","created_at":"2026-07-05T04:12:28.443072+00:00"},{"alias_kind":"pith_short_16","alias_value":"RSGS7RD25ZYA5DB3","created_at":"2026-07-05T04:12:28.443072+00:00"},{"alias_kind":"pith_short_8","alias_value":"RSGS7RD2","created_at":"2026-07-05T04:12:28.443072+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.29805","citing_title":"Clearer Sight, Fewer Lies: Oriented Pickup Preference Optimization for Multimodal Hallucination Mitigation","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2606.29805","citing_title":"Clearer Sight, Fewer Lies: Oriented Pickup Preference Optimization for Multimodal Hallucination Mitigation","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF","json":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF.json","graph_json":"https://pith.science/api/pith-number/RSGS7RD25ZYA5DB3KSOFL32NJF/graph.json","events_json":"https://pith.science/api/pith-number/RSGS7RD25ZYA5DB3KSOFL32NJF/events.json","paper":"https://pith.science/paper/RSGS7RD2"},"agent_actions":{"view_html":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF","download_json":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF.json","view_paper":"https://pith.science/paper/RSGS7RD2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.04121&json=true","fetch_graph":"https://pith.science/api/pith-number/RSGS7RD25ZYA5DB3KSOFL32NJF/graph.json","fetch_events":"https://pith.science/api/pith-number/RSGS7RD25ZYA5DB3KSOFL32NJF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF/action/storage_attestation","attest_author":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF/action/author_attestation","sign_citation":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF/action/citation_signature","submit_replication":"https://pith.science/pith/RSGS7RD25ZYA5DB3KSOFL32NJF/action/replication_record"}},"created_at":"2026-07-05T04:12:28.443072+00:00","updated_at":"2026-07-05T04:12:28.443072+00:00"}