{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:TF6IPNIFWG4WS6FTHNDWJGND36","short_pith_number":"pith:TF6IPNIF","schema_version":"1.0","canonical_sha256":"997c87b505b1b96978b33b476499a3dfb2303c403e0455bf45ec12f52c5494f3","source":{"kind":"arxiv","id":"2110.15154","version":1},"attestation_state":"computed","paper":{"title":"Cross-Batch Negative Sampling for Training Two-Tower Recommenders","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.IR","authors_text":"Jieming Zhu, Jinpeng Wang, Xiuqiang He","submitted_at":"2021-10-28T14:26:42Z","abstract_excerpt":"The two-tower architecture has been widely applied for learning item and user representations, which is important for large-scale recommender systems. Many two-tower models are trained using various in-batch negative sampling strategies, where the effects of such strategies inherently rely on the size of mini-batches. However, training two-tower models with a large batch size is inefficient, as it demands a large volume of memory for item and user contents and consumes a lot of time for feature encoding. Interestingly, we find that neural encoders can output relatively stable features for the "},"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":"2110.15154","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.IR","submitted_at":"2021-10-28T14:26:42Z","cross_cats_sorted":[],"title_canon_sha256":"376935ad46a4d844f5d9f9075fbdf86f885e9bad02d8faa525615fb2b2bc3a85","abstract_canon_sha256":"e8e1394ec18ce75609dbe7938349ab58a606f1625a7a5a55bd3845b2a1f24d03"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:26:52.279097Z","signature_b64":"qjekNyWgKKUx34cbnSKOt7dij3bCJBqyoiGmRa1cQYbWftzjgjnJgK9d/95sEPk7eTQN3eotfp/76yac5ka8Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"997c87b505b1b96978b33b476499a3dfb2303c403e0455bf45ec12f52c5494f3","last_reissued_at":"2026-07-05T03:26:52.278753Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:26:52.278753Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cross-Batch Negative Sampling for Training Two-Tower Recommenders","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.IR","authors_text":"Jieming Zhu, Jinpeng Wang, Xiuqiang He","submitted_at":"2021-10-28T14:26:42Z","abstract_excerpt":"The two-tower architecture has been widely applied for learning item and user representations, which is important for large-scale recommender systems. Many two-tower models are trained using various in-batch negative sampling strategies, where the effects of such strategies inherently rely on the size of mini-batches. However, training two-tower models with a large batch size is inefficient, as it demands a large volume of memory for item and user contents and consumes a lot of time for feature encoding. Interestingly, we find that neural encoders can output relatively stable features for the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.15154","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/2110.15154/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":"2110.15154","created_at":"2026-07-05T03:26:52.278813+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.15154v1","created_at":"2026-07-05T03:26:52.278813+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.15154","created_at":"2026-07-05T03:26:52.278813+00:00"},{"alias_kind":"pith_short_12","alias_value":"TF6IPNIFWG4W","created_at":"2026-07-05T03:26:52.278813+00:00"},{"alias_kind":"pith_short_16","alias_value":"TF6IPNIFWG4WS6FT","created_at":"2026-07-05T03:26:52.278813+00:00"},{"alias_kind":"pith_short_8","alias_value":"TF6IPNIF","created_at":"2026-07-05T03:26:52.278813+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/TF6IPNIFWG4WS6FTHNDWJGND36","json":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36.json","graph_json":"https://pith.science/api/pith-number/TF6IPNIFWG4WS6FTHNDWJGND36/graph.json","events_json":"https://pith.science/api/pith-number/TF6IPNIFWG4WS6FTHNDWJGND36/events.json","paper":"https://pith.science/paper/TF6IPNIF"},"agent_actions":{"view_html":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36","download_json":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36.json","view_paper":"https://pith.science/paper/TF6IPNIF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.15154&json=true","fetch_graph":"https://pith.science/api/pith-number/TF6IPNIFWG4WS6FTHNDWJGND36/graph.json","fetch_events":"https://pith.science/api/pith-number/TF6IPNIFWG4WS6FTHNDWJGND36/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36/action/storage_attestation","attest_author":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36/action/author_attestation","sign_citation":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36/action/citation_signature","submit_replication":"https://pith.science/pith/TF6IPNIFWG4WS6FTHNDWJGND36/action/replication_record"}},"created_at":"2026-07-05T03:26:52.278813+00:00","updated_at":"2026-07-05T03:26:52.278813+00:00"}