{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BLJI2TYTM4BDQNY5MJHSDD7ZO3","short_pith_number":"pith:BLJI2TYT","schema_version":"1.0","canonical_sha256":"0ad28d4f13670238371d624f218ff976c91f2e88506e565c7bbe0c2188b1e269","source":{"kind":"arxiv","id":"2505.15417","version":1},"attestation_state":"computed","paper":{"title":"Robust Multimodal Learning via Entropy-Gated Contrastive Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Leon Chlon, Maggie Chlon, MarcAntonio M. Awada","submitted_at":"2025-05-21T12:00:37Z","abstract_excerpt":"Real-world multimodal systems routinely face missing-input scenarios, and in reality, robots lose audio in a factory or a clinical record omits lab tests at inference time. Standard fusion layers either preserve robustness or calibration but never both. We introduce Adaptive Entropy-Gated Contrastive Fusion (AECF), a single light-weight layer that (i) adapts its entropy coefficient per instance, (ii) enforces monotone calibration across all modality subsets, and (iii) drives a curriculum mask directly from training-time entropy. On AV-MNIST and MS-COCO, AECF improves masked-input mAP by +18 pp"},"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":"2505.15417","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2025-05-21T12:00:37Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"8d08a5daf19793b22e9a8ad533de51a5f7e8aa5e9a51b66f847b63ff5c93fb65","abstract_canon_sha256":"208377d89348d2796d8dd65ec0fd1f56190b963396b050503d66f13a5f32cc52"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:45.248753Z","signature_b64":"nodIhPBs3YkKVnXQkRi0dcGrpPHrJNnlE6Y4DZE1CBqMtkAi86HDRRkOiilGcTpeMcPxceVyoNLGyb1EGrPDAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0ad28d4f13670238371d624f218ff976c91f2e88506e565c7bbe0c2188b1e269","last_reissued_at":"2026-07-05T11:06:45.248344Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:45.248344Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Robust Multimodal Learning via Entropy-Gated Contrastive Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Leon Chlon, Maggie Chlon, MarcAntonio M. Awada","submitted_at":"2025-05-21T12:00:37Z","abstract_excerpt":"Real-world multimodal systems routinely face missing-input scenarios, and in reality, robots lose audio in a factory or a clinical record omits lab tests at inference time. Standard fusion layers either preserve robustness or calibration but never both. We introduce Adaptive Entropy-Gated Contrastive Fusion (AECF), a single light-weight layer that (i) adapts its entropy coefficient per instance, (ii) enforces monotone calibration across all modality subsets, and (iii) drives a curriculum mask directly from training-time entropy. On AV-MNIST and MS-COCO, AECF improves masked-input mAP by +18 pp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.15417","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/2505.15417/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":"2505.15417","created_at":"2026-07-05T11:06:45.248404+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.15417v1","created_at":"2026-07-05T11:06:45.248404+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.15417","created_at":"2026-07-05T11:06:45.248404+00:00"},{"alias_kind":"pith_short_12","alias_value":"BLJI2TYTM4BD","created_at":"2026-07-05T11:06:45.248404+00:00"},{"alias_kind":"pith_short_16","alias_value":"BLJI2TYTM4BDQNY5","created_at":"2026-07-05T11:06:45.248404+00:00"},{"alias_kind":"pith_short_8","alias_value":"BLJI2TYT","created_at":"2026-07-05T11:06:45.248404+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.31135","citing_title":"MSNN-LINet: Cross-Modal Learning via Continuous Linear Integration","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3","json":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3.json","graph_json":"https://pith.science/api/pith-number/BLJI2TYTM4BDQNY5MJHSDD7ZO3/graph.json","events_json":"https://pith.science/api/pith-number/BLJI2TYTM4BDQNY5MJHSDD7ZO3/events.json","paper":"https://pith.science/paper/BLJI2TYT"},"agent_actions":{"view_html":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3","download_json":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3.json","view_paper":"https://pith.science/paper/BLJI2TYT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.15417&json=true","fetch_graph":"https://pith.science/api/pith-number/BLJI2TYTM4BDQNY5MJHSDD7ZO3/graph.json","fetch_events":"https://pith.science/api/pith-number/BLJI2TYTM4BDQNY5MJHSDD7ZO3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3/action/storage_attestation","attest_author":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3/action/author_attestation","sign_citation":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3/action/citation_signature","submit_replication":"https://pith.science/pith/BLJI2TYTM4BDQNY5MJHSDD7ZO3/action/replication_record"}},"created_at":"2026-07-05T11:06:45.248404+00:00","updated_at":"2026-07-05T11:06:45.248404+00:00"}