{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JRB2HAAGM57ATG2NC3KBIP4BCZ","short_pith_number":"pith:JRB2HAAG","schema_version":"1.0","canonical_sha256":"4c43a38006677e099b4d16d4143f811646b3c5691e20c4b3bf2a4eae851835f6","source":{"kind":"arxiv","id":"2607.07135","version":1},"attestation_state":"computed","paper":{"title":"Sparse Attention for Dense Open-Vocabulary Prediction in CLIP","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bernard Ghanem, Chen Zhao, Fatimah Zohra, Shuming Liu","submitted_at":"2026-07-08T08:30:48Z","abstract_excerpt":"Contrastive Language-Image Pre-training (CLIP) relies on softmax-based self-attention, a strictly positive distribution that assigns probability mass to every pair of tokens-even semantically irrelevant ones. While these dense softmax weights are effective for gathering broad context during pre-training, they spread attention across many low-salience tokens, producing noise that obscures the fine-grained, spatially localized cues required for dense, open-vocabulary prediction. We study an inference-time substitution of the row-wise softmax in the final visual self-attention layers with the $\\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":"2607.07135","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-08T08:30:48Z","cross_cats_sorted":[],"title_canon_sha256":"7ae019c46dfed7a7c2c06ca7f46256cf95a18953398e2627fe962f0102b9ac55","abstract_canon_sha256":"3b510813fce759e586eaa14c4c10380ca0d125b5703c5448a9221d028af7be04"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-09T01:20:15.487185Z","signature_b64":"2IDFwJ5y9USdLfVich1IGI5nYnj7RnHTTUzu5ecjKTpoF9gLOJpOpSs6NoKH9KbdTDLEoEMfiFPPmhTYBXCPAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4c43a38006677e099b4d16d4143f811646b3c5691e20c4b3bf2a4eae851835f6","last_reissued_at":"2026-07-09T01:20:15.486806Z","signature_status":"signed_v1","first_computed_at":"2026-07-09T01:20:15.486806Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sparse Attention for Dense Open-Vocabulary Prediction in CLIP","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Bernard Ghanem, Chen Zhao, Fatimah Zohra, Shuming Liu","submitted_at":"2026-07-08T08:30:48Z","abstract_excerpt":"Contrastive Language-Image Pre-training (CLIP) relies on softmax-based self-attention, a strictly positive distribution that assigns probability mass to every pair of tokens-even semantically irrelevant ones. While these dense softmax weights are effective for gathering broad context during pre-training, they spread attention across many low-salience tokens, producing noise that obscures the fine-grained, spatially localized cues required for dense, open-vocabulary prediction. We study an inference-time substitution of the row-wise softmax in the final visual self-attention layers with the $\\a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.07135","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/2607.07135/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":"2607.07135","created_at":"2026-07-09T01:20:15.486861+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.07135v1","created_at":"2026-07-09T01:20:15.486861+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.07135","created_at":"2026-07-09T01:20:15.486861+00:00"},{"alias_kind":"pith_short_12","alias_value":"JRB2HAAGM57A","created_at":"2026-07-09T01:20:15.486861+00:00"},{"alias_kind":"pith_short_16","alias_value":"JRB2HAAGM57ATG2N","created_at":"2026-07-09T01:20:15.486861+00:00"},{"alias_kind":"pith_short_8","alias_value":"JRB2HAAG","created_at":"2026-07-09T01:20:15.486861+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/JRB2HAAGM57ATG2NC3KBIP4BCZ","json":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ.json","graph_json":"https://pith.science/api/pith-number/JRB2HAAGM57ATG2NC3KBIP4BCZ/graph.json","events_json":"https://pith.science/api/pith-number/JRB2HAAGM57ATG2NC3KBIP4BCZ/events.json","paper":"https://pith.science/paper/JRB2HAAG"},"agent_actions":{"view_html":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ","download_json":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ.json","view_paper":"https://pith.science/paper/JRB2HAAG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.07135&json=true","fetch_graph":"https://pith.science/api/pith-number/JRB2HAAGM57ATG2NC3KBIP4BCZ/graph.json","fetch_events":"https://pith.science/api/pith-number/JRB2HAAGM57ATG2NC3KBIP4BCZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ/action/storage_attestation","attest_author":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ/action/author_attestation","sign_citation":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ/action/citation_signature","submit_replication":"https://pith.science/pith/JRB2HAAGM57ATG2NC3KBIP4BCZ/action/replication_record"}},"created_at":"2026-07-09T01:20:15.486861+00:00","updated_at":"2026-07-09T01:20:15.486861+00:00"}