{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:RWWQ6VDL3UNKMRYJ4EMO4NLMWR","short_pith_number":"pith:RWWQ6VDL","schema_version":"1.0","canonical_sha256":"8dad0f546bdd1aa64709e118ee356cb449f575ab90e442f4333f581df5bf0e48","source":{"kind":"arxiv","id":"2006.05744","version":2},"attestation_state":"computed","paper":{"title":"MC-BERT: Efficient Language Pre-Training via a Meta Controller","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Di He, Guolin Ke, Jiang Bian, Linyuan Gong, Liwei Wang, Shuxin Zheng, Tie-Yan Liu, Zhenhui Xu","submitted_at":"2020-06-10T09:22:19Z","abstract_excerpt":"Pre-trained contextual representations (e.g., BERT) have become the foundation to achieve state-of-the-art results on many NLP tasks. However, large-scale pre-training is computationally expensive. ELECTRA, an early attempt to accelerate pre-training, trains a discriminative model that predicts whether each input token was replaced by a generator. Our studies reveal that ELECTRA's success is mainly due to its reduced complexity of the pre-training task: the binary classification (replaced token detection) is more efficient to learn than the generation task (masked language modeling). However, "},"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":"2006.05744","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-06-10T09:22:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"69dc6a64a836e383d98646f63cd0a436c13de97bc2dc3e5ec02306d6aa63dc70","abstract_canon_sha256":"f76cb6b7c39d4d021240efab7517f99a616927d09e24efe267f7dd59813f167c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:10:40.630676Z","signature_b64":"/IZDYGMamIq6obtNS4gfYypmJ/4KrcgZZr/e4lzVYGgEQJVsWi4WtS37GlekGbpjCAb+Mse5yxCqVq7YV2HDCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8dad0f546bdd1aa64709e118ee356cb449f575ab90e442f4333f581df5bf0e48","last_reissued_at":"2026-07-05T01:10:40.630124Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:10:40.630124Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MC-BERT: Efficient Language Pre-Training via a Meta Controller","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Di He, Guolin Ke, Jiang Bian, Linyuan Gong, Liwei Wang, Shuxin Zheng, Tie-Yan Liu, Zhenhui Xu","submitted_at":"2020-06-10T09:22:19Z","abstract_excerpt":"Pre-trained contextual representations (e.g., BERT) have become the foundation to achieve state-of-the-art results on many NLP tasks. However, large-scale pre-training is computationally expensive. ELECTRA, an early attempt to accelerate pre-training, trains a discriminative model that predicts whether each input token was replaced by a generator. Our studies reveal that ELECTRA's success is mainly due to its reduced complexity of the pre-training task: the binary classification (replaced token detection) is more efficient to learn than the generation task (masked language modeling). However, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2006.05744","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/2006.05744/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":"2006.05744","created_at":"2026-07-05T01:10:40.630195+00:00"},{"alias_kind":"arxiv_version","alias_value":"2006.05744v2","created_at":"2026-07-05T01:10:40.630195+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2006.05744","created_at":"2026-07-05T01:10:40.630195+00:00"},{"alias_kind":"pith_short_12","alias_value":"RWWQ6VDL3UNK","created_at":"2026-07-05T01:10:40.630195+00:00"},{"alias_kind":"pith_short_16","alias_value":"RWWQ6VDL3UNKMRYJ","created_at":"2026-07-05T01:10:40.630195+00:00"},{"alias_kind":"pith_short_8","alias_value":"RWWQ6VDL","created_at":"2026-07-05T01:10:40.630195+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22133","citing_title":"Atom-level Protein Representation Learning Improves Protein Structure Prediction","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22133","citing_title":"Atom-level Protein Representation Learning Improves Protein Structure Prediction","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22133","citing_title":"Atom-level Protein Representation Learning Improves Protein Structure Prediction","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03103","citing_title":"MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR","json":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR.json","graph_json":"https://pith.science/api/pith-number/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/graph.json","events_json":"https://pith.science/api/pith-number/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/events.json","paper":"https://pith.science/paper/RWWQ6VDL"},"agent_actions":{"view_html":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR","download_json":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR.json","view_paper":"https://pith.science/paper/RWWQ6VDL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2006.05744&json=true","fetch_graph":"https://pith.science/api/pith-number/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/graph.json","fetch_events":"https://pith.science/api/pith-number/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/action/storage_attestation","attest_author":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/action/author_attestation","sign_citation":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/action/citation_signature","submit_replication":"https://pith.science/pith/RWWQ6VDL3UNKMRYJ4EMO4NLMWR/action/replication_record"}},"created_at":"2026-07-05T01:10:40.630195+00:00","updated_at":"2026-07-05T01:10:40.630195+00:00"}