{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:7AQOW4YKPJ3WM75FAUIJMK7C5L","short_pith_number":"pith:7AQOW4YK","schema_version":"1.0","canonical_sha256":"f820eb730a7a77667fa50510962be2eaed327734d0aa70ff3623bb72a0b2469d","source":{"kind":"arxiv","id":"2205.12035","version":2},"attestation_state":"computed","paper":{"title":"RetroMAE: Pre-Training Retrieval-oriented Language Models Via Masked Auto-Encoder","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Shitao Xiao, Yingxia Shao, Zhao Cao, Zheng Liu","submitted_at":"2022-05-24T12:43:04Z","abstract_excerpt":"Despite pre-training's progress in many important NLP tasks, it remains to explore effective pre-training strategies for dense retrieval. In this paper, we propose RetroMAE, a new retrieval oriented pre-training paradigm based on Masked Auto-Encoder (MAE). RetroMAE is highlighted by three critical designs. 1) A novel MAE workflow, where the input sentence is polluted for encoder and decoder with different masks. The sentence embedding is generated from the encoder's masked input; then, the original sentence is recovered based on the sentence embedding and the decoder's masked input via masked "},"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":"2205.12035","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-05-24T12:43:04Z","cross_cats_sorted":[],"title_canon_sha256":"abe77ddc0d0006f4af3125b2f7be382b821d419015b038c65976a88fdc233ff1","abstract_canon_sha256":"57fb9e7b0b4edfc054233a3769fb8fa421336c7980117b104ba2ca5c8c93efd0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:06:58.553126Z","signature_b64":"NHD7IPSDG0ZNubSX9SbhQq8zMG4ndFvAQqVDu6uCJeG3SWo1zZdbH+xgjt9mGE1KMwMALuxVGFGkBiaV5mnABQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f820eb730a7a77667fa50510962be2eaed327734d0aa70ff3623bb72a0b2469d","last_reissued_at":"2026-07-05T05:06:58.552498Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:06:58.552498Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RetroMAE: Pre-Training Retrieval-oriented Language Models Via Masked Auto-Encoder","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Shitao Xiao, Yingxia Shao, Zhao Cao, Zheng Liu","submitted_at":"2022-05-24T12:43:04Z","abstract_excerpt":"Despite pre-training's progress in many important NLP tasks, it remains to explore effective pre-training strategies for dense retrieval. In this paper, we propose RetroMAE, a new retrieval oriented pre-training paradigm based on Masked Auto-Encoder (MAE). RetroMAE is highlighted by three critical designs. 1) A novel MAE workflow, where the input sentence is polluted for encoder and decoder with different masks. The sentence embedding is generated from the encoder's masked input; then, the original sentence is recovered based on the sentence embedding and the decoder's masked input via masked "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2205.12035","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/2205.12035/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":"2205.12035","created_at":"2026-07-05T05:06:58.552581+00:00"},{"alias_kind":"arxiv_version","alias_value":"2205.12035v2","created_at":"2026-07-05T05:06:58.552581+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2205.12035","created_at":"2026-07-05T05:06:58.552581+00:00"},{"alias_kind":"pith_short_12","alias_value":"7AQOW4YKPJ3W","created_at":"2026-07-05T05:06:58.552581+00:00"},{"alias_kind":"pith_short_16","alias_value":"7AQOW4YKPJ3WM75F","created_at":"2026-07-05T05:06:58.552581+00:00"},{"alias_kind":"pith_short_8","alias_value":"7AQOW4YK","created_at":"2026-07-05T05:06:58.552581+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2309.07597","citing_title":"C-Pack: Packed Resources For General Chinese Embeddings","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27037","citing_title":"Hypencoder Revisited: Reproducibility and Analysis of Non-Linear Scoring for First-Stage Retrieval","ref_index":67,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10937","citing_title":"Benchmarking and Enabling Efficient Chinese Medical Retrieval via Asymmetric Encoders","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06163","citing_title":"Data, Not Model: Explaining Bias toward LLM Texts in Neural Retrievers","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L","json":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L.json","graph_json":"https://pith.science/api/pith-number/7AQOW4YKPJ3WM75FAUIJMK7C5L/graph.json","events_json":"https://pith.science/api/pith-number/7AQOW4YKPJ3WM75FAUIJMK7C5L/events.json","paper":"https://pith.science/paper/7AQOW4YK"},"agent_actions":{"view_html":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L","download_json":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L.json","view_paper":"https://pith.science/paper/7AQOW4YK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2205.12035&json=true","fetch_graph":"https://pith.science/api/pith-number/7AQOW4YKPJ3WM75FAUIJMK7C5L/graph.json","fetch_events":"https://pith.science/api/pith-number/7AQOW4YKPJ3WM75FAUIJMK7C5L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L/action/storage_attestation","attest_author":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L/action/author_attestation","sign_citation":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L/action/citation_signature","submit_replication":"https://pith.science/pith/7AQOW4YKPJ3WM75FAUIJMK7C5L/action/replication_record"}},"created_at":"2026-07-05T05:06:58.552581+00:00","updated_at":"2026-07-05T05:06:58.552581+00:00"}