{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:UOEYCMKSCMVHMHK7LQSP7KSEE6","short_pith_number":"pith:UOEYCMKS","schema_version":"1.0","canonical_sha256":"a389813152132a761d5f5c24ffaa4427b89731ae544d5f8524b70c569a78dee2","source":{"kind":"arxiv","id":"2204.06644","version":2},"attestation_state":"computed","paper":{"title":"METRO: Efficient Denoising Pretraining of Large Scale Autoencoding Language Models with Model Generated Signals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Chenyan Xiong, Di He, Guolin Ke, Jianfeng Gao, Paul Bennett, Payal Bajaj, Saurabh Tiwary, Tie-Yan Liu, XiaoDong Liu, Xia Song","submitted_at":"2022-04-13T21:39:15Z","abstract_excerpt":"We present an efficient method of pretraining large-scale autoencoding language models using training signals generated by an auxiliary model. Originated in ELECTRA, this training strategy has demonstrated sample-efficiency to pretrain models at the scale of hundreds of millions of parameters. In this work, we conduct a comprehensive empirical study, and propose a recipe, namely \"Model generated dEnoising TRaining Objective\" (METRO), which incorporates some of the best modeling techniques developed recently to speed up, stabilize, and enhance pretrained language models without compromising mod"},"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":"2204.06644","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-04-13T21:39:15Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"a9f27c19e3333562bcca50fb6f207c88ede9b05807d871e729d75faa12a730cf","abstract_canon_sha256":"3792bece98dcf1dc834b77b7d01b05fb70ad27b49f2fd184d08e88a8d5406b12"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:15:16.088491Z","signature_b64":"B6/Rv6JVwfTCLPZzrtml8qBv/HQpsYv7WwDOWpjFnvVItatS1WE8hS/EcRcN8EKfS3k+VyUKLXujcs3wLBcUCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a389813152132a761d5f5c24ffaa4427b89731ae544d5f8524b70c569a78dee2","last_reissued_at":"2026-07-05T04:15:16.088042Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:15:16.088042Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"METRO: Efficient Denoising Pretraining of Large Scale Autoencoding Language Models with Model Generated Signals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Chenyan Xiong, Di He, Guolin Ke, Jianfeng Gao, Paul Bennett, Payal Bajaj, Saurabh Tiwary, Tie-Yan Liu, XiaoDong Liu, Xia Song","submitted_at":"2022-04-13T21:39:15Z","abstract_excerpt":"We present an efficient method of pretraining large-scale autoencoding language models using training signals generated by an auxiliary model. Originated in ELECTRA, this training strategy has demonstrated sample-efficiency to pretrain models at the scale of hundreds of millions of parameters. In this work, we conduct a comprehensive empirical study, and propose a recipe, namely \"Model generated dEnoising TRaining Objective\" (METRO), which incorporates some of the best modeling techniques developed recently to speed up, stabilize, and enhance pretrained language models without compromising mod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2204.06644","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/2204.06644/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":"2204.06644","created_at":"2026-07-05T04:15:16.088099+00:00"},{"alias_kind":"arxiv_version","alias_value":"2204.06644v2","created_at":"2026-07-05T04:15:16.088099+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2204.06644","created_at":"2026-07-05T04:15:16.088099+00:00"},{"alias_kind":"pith_short_12","alias_value":"UOEYCMKSCMVH","created_at":"2026-07-05T04:15:16.088099+00:00"},{"alias_kind":"pith_short_16","alias_value":"UOEYCMKSCMVHMHK7","created_at":"2026-07-05T04:15:16.088099+00:00"},{"alias_kind":"pith_short_8","alias_value":"UOEYCMKS","created_at":"2026-07-05T04:15:16.088099+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2310.17591","citing_title":"Lil-Bevo: Explorations of Strategies for Training Language Models in More Humanlike Ways","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6","json":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6.json","graph_json":"https://pith.science/api/pith-number/UOEYCMKSCMVHMHK7LQSP7KSEE6/graph.json","events_json":"https://pith.science/api/pith-number/UOEYCMKSCMVHMHK7LQSP7KSEE6/events.json","paper":"https://pith.science/paper/UOEYCMKS"},"agent_actions":{"view_html":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6","download_json":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6.json","view_paper":"https://pith.science/paper/UOEYCMKS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2204.06644&json=true","fetch_graph":"https://pith.science/api/pith-number/UOEYCMKSCMVHMHK7LQSP7KSEE6/graph.json","fetch_events":"https://pith.science/api/pith-number/UOEYCMKSCMVHMHK7LQSP7KSEE6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6/action/storage_attestation","attest_author":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6/action/author_attestation","sign_citation":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6/action/citation_signature","submit_replication":"https://pith.science/pith/UOEYCMKSCMVHMHK7LQSP7KSEE6/action/replication_record"}},"created_at":"2026-07-05T04:15:16.088099+00:00","updated_at":"2026-07-05T04:15:16.088099+00:00"}