{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:2FOIZVO2544P5CKF3BT6BSZPMA","short_pith_number":"pith:2FOIZVO2","schema_version":"1.0","canonical_sha256":"d15c8cd5daef38fe8945d867e0cb2f60300a1eba5028ff3d5d57106b61ebf8ca","source":{"kind":"arxiv","id":"2607.17673","version":1},"attestation_state":"computed","paper":{"title":"Beyond Objective Expressivity: Geometry Preservation in Multimodal Contrastive Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Benjamin Wild, Roland Eils, Tillmann Rheude","submitted_at":"2026-07-20T08:25:45Z","abstract_excerpt":"Contrastive learning is increasingly moving toward settings with three or more modalities instead of image-text pairs. Yet, extending models from pairwise to higher-order multimodal alignment can introduce optimization and representation challenges. We identify encoder Jacobian conditioning as a key factor in trimodal contrastive learning: poorly conditioned encoders exhibit collapsing or amplified singular-value spectra, leading to exploding Jacobian condition numbers and degraded multimodal alignment. We introduce geometry-preserving encoders (GPEs) by directly conditioning the Jacobian thro"},"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":"2607.17673","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-20T08:25:45Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"df1cfed07106be68f2ddbeb2de88ec9b3971156898566eaaa52636f75087fa56","abstract_canon_sha256":"90945e195b183ddf6b284fd477ba6666facf04b3d01d1295c022413a7c460238"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T02:21:53.918031Z","signature_b64":"XvotmIH4hh+BkXfv8qRqkYZ8WLSuUK7XFxU0Nc2U8fnhEL0l+tMFypH8ce13DBgxnP5Crd+l40Yb1hSbYasgCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d15c8cd5daef38fe8945d867e0cb2f60300a1eba5028ff3d5d57106b61ebf8ca","last_reissued_at":"2026-07-21T02:21:53.917199Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T02:21:53.917199Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Objective Expressivity: Geometry Preservation in Multimodal Contrastive Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Benjamin Wild, Roland Eils, Tillmann Rheude","submitted_at":"2026-07-20T08:25:45Z","abstract_excerpt":"Contrastive learning is increasingly moving toward settings with three or more modalities instead of image-text pairs. Yet, extending models from pairwise to higher-order multimodal alignment can introduce optimization and representation challenges. We identify encoder Jacobian conditioning as a key factor in trimodal contrastive learning: poorly conditioned encoders exhibit collapsing or amplified singular-value spectra, leading to exploding Jacobian condition numbers and degraded multimodal alignment. We introduce geometry-preserving encoders (GPEs) by directly conditioning the Jacobian thro"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.17673","kind":"arxiv","version":1},"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/2607.17673/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":"2607.17673","created_at":"2026-07-21T02:21:53.917642+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.17673v1","created_at":"2026-07-21T02:21:53.917642+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.17673","created_at":"2026-07-21T02:21:53.917642+00:00"},{"alias_kind":"pith_short_12","alias_value":"2FOIZVO2544P","created_at":"2026-07-21T02:21:53.917642+00:00"},{"alias_kind":"pith_short_16","alias_value":"2FOIZVO2544P5CKF","created_at":"2026-07-21T02:21:53.917642+00:00"},{"alias_kind":"pith_short_8","alias_value":"2FOIZVO2","created_at":"2026-07-21T02:21:53.917642+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA","json":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA.json","graph_json":"https://pith.science/api/pith-number/2FOIZVO2544P5CKF3BT6BSZPMA/graph.json","events_json":"https://pith.science/api/pith-number/2FOIZVO2544P5CKF3BT6BSZPMA/events.json","paper":"https://pith.science/paper/2FOIZVO2"},"agent_actions":{"view_html":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA","download_json":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA.json","view_paper":"https://pith.science/paper/2FOIZVO2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.17673&json=true","fetch_graph":"https://pith.science/api/pith-number/2FOIZVO2544P5CKF3BT6BSZPMA/graph.json","fetch_events":"https://pith.science/api/pith-number/2FOIZVO2544P5CKF3BT6BSZPMA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA/action/storage_attestation","attest_author":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA/action/author_attestation","sign_citation":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA/action/citation_signature","submit_replication":"https://pith.science/pith/2FOIZVO2544P5CKF3BT6BSZPMA/action/replication_record"}},"created_at":"2026-07-21T02:21:53.917642+00:00","updated_at":"2026-07-21T02:21:53.917642+00:00"}