{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:SIQEBBNC2FM76TNPAGFD3LJPWH","short_pith_number":"pith:SIQEBBNC","schema_version":"1.0","canonical_sha256":"92204085a2d159ff4daf018a3dad2fb1fa64698c8a26a6f0095af577befd7e17","source":{"kind":"arxiv","id":"2403.02875","version":2},"attestation_state":"computed","paper":{"title":"Enhancing Conceptual Understanding in Multimodal Contrastive Learning through Hard Negative Samples","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.IR"],"primary_cat":"cs.CV","authors_text":"Jind\\v{r}ich Libovick\\'y, Michaela Geierhos, Norbert Oswald, Philipp J. R\\\"osch","submitted_at":"2024-03-05T11:38:48Z","abstract_excerpt":"Current multimodal models leveraging contrastive learning often face limitations in developing fine-grained conceptual understanding. This is due to random negative samples during pretraining, causing almost exclusively very dissimilar concepts to be compared in the loss function. Consequently, the models struggle with fine-grained semantic differences. To address this problem, we introduce a novel pretraining method incorporating synthetic hard negative text examples. The hard negatives permute terms corresponding to visual concepts, leading to a more fine-grained visual and textual concept a"},"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":"2403.02875","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-05T11:38:48Z","cross_cats_sorted":["cs.CL","cs.IR"],"title_canon_sha256":"61df1d03610b1ac134334fbbd33756ba8b4c37eb3224df5a1b463437ccc2e6e4","abstract_canon_sha256":"2c68bbedd8d840cdd5c8d8c9053b43fa4349150752f4063e0938bf3f1c7eb574"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:51:48.980188Z","signature_b64":"xpo/ZS0K8nRibsgJVt/z5wjS/g1lSmI0Nv/eTgf3ekx/LFrj9Gec1s/c0l/uci5Zaayq8VVPB8WwVwBOFixFDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"92204085a2d159ff4daf018a3dad2fb1fa64698c8a26a6f0095af577befd7e17","last_reissued_at":"2026-07-05T08:51:48.979753Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:51:48.979753Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Enhancing Conceptual Understanding in Multimodal Contrastive Learning through Hard Negative Samples","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.IR"],"primary_cat":"cs.CV","authors_text":"Jind\\v{r}ich Libovick\\'y, Michaela Geierhos, Norbert Oswald, Philipp J. R\\\"osch","submitted_at":"2024-03-05T11:38:48Z","abstract_excerpt":"Current multimodal models leveraging contrastive learning often face limitations in developing fine-grained conceptual understanding. This is due to random negative samples during pretraining, causing almost exclusively very dissimilar concepts to be compared in the loss function. Consequently, the models struggle with fine-grained semantic differences. To address this problem, we introduce a novel pretraining method incorporating synthetic hard negative text examples. The hard negatives permute terms corresponding to visual concepts, leading to a more fine-grained visual and textual concept a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.02875","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/2403.02875/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":"2403.02875","created_at":"2026-07-05T08:51:48.979811+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.02875v2","created_at":"2026-07-05T08:51:48.979811+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.02875","created_at":"2026-07-05T08:51:48.979811+00:00"},{"alias_kind":"pith_short_12","alias_value":"SIQEBBNC2FM7","created_at":"2026-07-05T08:51:48.979811+00:00"},{"alias_kind":"pith_short_16","alias_value":"SIQEBBNC2FM76TNP","created_at":"2026-07-05T08:51:48.979811+00:00"},{"alias_kind":"pith_short_8","alias_value":"SIQEBBNC","created_at":"2026-07-05T08:51:48.979811+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.12748","citing_title":"Scaling Representation Diversity: Modulated Attention and Reconstructive Regularization for Visual Grounding","ref_index":69,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH","json":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH.json","graph_json":"https://pith.science/api/pith-number/SIQEBBNC2FM76TNPAGFD3LJPWH/graph.json","events_json":"https://pith.science/api/pith-number/SIQEBBNC2FM76TNPAGFD3LJPWH/events.json","paper":"https://pith.science/paper/SIQEBBNC"},"agent_actions":{"view_html":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH","download_json":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH.json","view_paper":"https://pith.science/paper/SIQEBBNC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.02875&json=true","fetch_graph":"https://pith.science/api/pith-number/SIQEBBNC2FM76TNPAGFD3LJPWH/graph.json","fetch_events":"https://pith.science/api/pith-number/SIQEBBNC2FM76TNPAGFD3LJPWH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH/action/storage_attestation","attest_author":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH/action/author_attestation","sign_citation":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH/action/citation_signature","submit_replication":"https://pith.science/pith/SIQEBBNC2FM76TNPAGFD3LJPWH/action/replication_record"}},"created_at":"2026-07-05T08:51:48.979811+00:00","updated_at":"2026-07-05T08:51:48.979811+00:00"}