{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:BGF34GYB2EWE42CL7LFZSA5QGO","short_pith_number":"pith:BGF34GYB","schema_version":"1.0","canonical_sha256":"098bbe1b01d12c4e684bfacb9903b033b4c2fa6a35aad1e9513d748c40d50f4d","source":{"kind":"arxiv","id":"2004.09297","version":2},"attestation_state":"computed","paper":{"title":"MPNet: Masked and Permuted Pre-training for Language Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Jianfeng Lu, Kaitao Song, Tao Qin, Tie-Yan Liu, Xu Tan","submitted_at":"2020-04-20T13:54:12Z","abstract_excerpt":"BERT adopts masked language modeling (MLM) for pre-training and is one of the most successful pre-training models. Since BERT neglects dependency among predicted tokens, XLNet introduces permuted language modeling (PLM) for pre-training to address this problem. However, XLNet does not leverage the full position information of a sentence and thus suffers from position discrepancy between pre-training and fine-tuning. In this paper, we propose MPNet, a novel pre-training method that inherits the advantages of BERT and XLNet and avoids their limitations. MPNet leverages the dependency among predi"},"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":"2004.09297","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-20T13:54:12Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ef3b8c51239f13e6bf7459363707b47782d57af41c975dd52bf81a7a121cb070","abstract_canon_sha256":"e23180ed8728568decb43e2c269e01284a62bc1c0db64306287cb9db97155b24"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:48:08.917264Z","signature_b64":"BateFMLwxHor/0lZX1x1qdbJYb0Rj3FMN3hsN4zxbLOvZ6fJ5b+gSVlaZwOxuqIEMnwjBRWfCNEvnvrH/5PcAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"098bbe1b01d12c4e684bfacb9903b033b4c2fa6a35aad1e9513d748c40d50f4d","last_reissued_at":"2026-07-05T01:48:08.916827Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:48:08.916827Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MPNet: Masked and Permuted Pre-training for Language Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Jianfeng Lu, Kaitao Song, Tao Qin, Tie-Yan Liu, Xu Tan","submitted_at":"2020-04-20T13:54:12Z","abstract_excerpt":"BERT adopts masked language modeling (MLM) for pre-training and is one of the most successful pre-training models. Since BERT neglects dependency among predicted tokens, XLNet introduces permuted language modeling (PLM) for pre-training to address this problem. However, XLNet does not leverage the full position information of a sentence and thus suffers from position discrepancy between pre-training and fine-tuning. In this paper, we propose MPNet, a novel pre-training method that inherits the advantages of BERT and XLNet and avoids their limitations. MPNet leverages the dependency among predi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.09297","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/2004.09297/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":"2004.09297","created_at":"2026-07-05T01:48:08.916883+00:00"},{"alias_kind":"arxiv_version","alias_value":"2004.09297v2","created_at":"2026-07-05T01:48:08.916883+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.09297","created_at":"2026-07-05T01:48:08.916883+00:00"},{"alias_kind":"pith_short_12","alias_value":"BGF34GYB2EWE","created_at":"2026-07-05T01:48:08.916883+00:00"},{"alias_kind":"pith_short_16","alias_value":"BGF34GYB2EWE42CL","created_at":"2026-07-05T01:48:08.916883+00:00"},{"alias_kind":"pith_short_8","alias_value":"BGF34GYB","created_at":"2026-07-05T01:48:08.916883+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25379","citing_title":"Story Operators: Decomposing the Original $\\to$ Sequel Transformation in Embedding Space","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26753","citing_title":"ConvMemory v3: A Validity Context Layer for Conversational Memory via Target-Conditioned Relation Verification","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16046","citing_title":"XSearch: Explainable Code Search via Concept-to-Code Alignment","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2607.01485","citing_title":"CoPersona: Collaborative Persona Graphs for Robust LLM Personalization","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2606.10842","citing_title":"ConvMemory v2: A Recall-Preserving Top-10 Evidence Reranker for Conversational Memory Retrieval","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09951","citing_title":"Hasse Diagrams for Attention: A Partial Order Framework for Designing Transformer Masks","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05436","citing_title":"Ten Headache Specialists versus Artificial Intelligence for Clinical Literature Summarization: A Critical Evaluation and Comparison","ref_index":81,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07587","citing_title":"The Routing Plateau: Understanding and Breaking the Accuracy Limits of LLM Routers","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28062","citing_title":"ConvMemory: A Lightweight Learned Memory Reranker, a Negative Attribution Result, and a Research-Preview Conflict Editor","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30729","citing_title":"SemStruct: Contextualizing Semantic Embeddings with Structural Information for Schema Matching","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2503.07878","citing_title":"A Woman with a Knife or A Knife with a Woman? Measuring Directional Bias Amplification in Image Captions","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21242","citing_title":"To Select or not to Select, that is the Question: Distilling Robot Skill Prediction into a Small Ensemble","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16046","citing_title":"XSearch: Explainable Code Search via Concept-to-Code Alignment","ref_index":71,"is_internal_anchor":false},{"citing_arxiv_id":"2508.16860","citing_title":"TriagerX: Dual Transformers for Bug Triaging Tasks with Content and Interaction Based Rankings","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06967","citing_title":"VulLink: A Dynamic Open-Access Vulnerability Graph Database for Cybersecurity Data Mining","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21555","citing_title":"Finding Meaning in Embeddings: Concept Separation Curves","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO","json":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO.json","graph_json":"https://pith.science/api/pith-number/BGF34GYB2EWE42CL7LFZSA5QGO/graph.json","events_json":"https://pith.science/api/pith-number/BGF34GYB2EWE42CL7LFZSA5QGO/events.json","paper":"https://pith.science/paper/BGF34GYB"},"agent_actions":{"view_html":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO","download_json":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO.json","view_paper":"https://pith.science/paper/BGF34GYB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2004.09297&json=true","fetch_graph":"https://pith.science/api/pith-number/BGF34GYB2EWE42CL7LFZSA5QGO/graph.json","fetch_events":"https://pith.science/api/pith-number/BGF34GYB2EWE42CL7LFZSA5QGO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO/action/storage_attestation","attest_author":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO/action/author_attestation","sign_citation":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO/action/citation_signature","submit_replication":"https://pith.science/pith/BGF34GYB2EWE42CL7LFZSA5QGO/action/replication_record"}},"created_at":"2026-07-05T01:48:08.916883+00:00","updated_at":"2026-07-05T01:48:08.916883+00:00"}