{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:D3PFMI63QW7LYDSFENT2OKXKFN","short_pith_number":"pith:D3PFMI63","schema_version":"1.0","canonical_sha256":"1ede5623db85bebc0e452367a72aea2b43c8e842bcbc727d0cc12f4b5861c867","source":{"kind":"arxiv","id":"2501.13397","version":5},"attestation_state":"computed","paper":{"title":"ExLM: Rethinking the Impact of [MASK] Tokens in Masked Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Bin Feng, Junwei Yang, Kangjie Zheng, Ming Zhang, Siyue Liang, Wei Ju, Zequn Liu, Zhiping Xiao","submitted_at":"2025-01-23T05:46:50Z","abstract_excerpt":"Masked Language Models (MLMs) have achieved remarkable success in many self-supervised representation learning tasks. MLMs are trained by randomly masking portions of the input sequences with [MASK] tokens and learning to reconstruct the original content based on the remaining context. This paper explores the impact of [MASK] tokens on MLMs. Analytical studies show that masking tokens can introduce the corrupted semantics problem, wherein the corrupted context may convey multiple, ambiguous meanings. This problem is also a key factor affecting the performance of MLMs on downstream tasks. Based"},"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":"2501.13397","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-23T05:46:50Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"38cf3faa035ce5f22f271ce90206ad42793b3d57a73e108db8c5a7bdb4cf4f9c","abstract_canon_sha256":"8e42c6f556c0712d1368535dacfa13f1cd79ce72a431748d749c7f2a03d5514c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:17:39.969130Z","signature_b64":"xEGaMxitaqpJ7BgfNXDGXNouw1BJAue9iOjazsaIWjrzzAP04ehWg4/Yw1ByVlhoylC8svU55u5qgLuqNf5pCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ede5623db85bebc0e452367a72aea2b43c8e842bcbc727d0cc12f4b5861c867","last_reissued_at":"2026-07-05T11:17:39.968648Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:17:39.968648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ExLM: Rethinking the Impact of [MASK] Tokens in Masked Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Bin Feng, Junwei Yang, Kangjie Zheng, Ming Zhang, Siyue Liang, Wei Ju, Zequn Liu, Zhiping Xiao","submitted_at":"2025-01-23T05:46:50Z","abstract_excerpt":"Masked Language Models (MLMs) have achieved remarkable success in many self-supervised representation learning tasks. MLMs are trained by randomly masking portions of the input sequences with [MASK] tokens and learning to reconstruct the original content based on the remaining context. This paper explores the impact of [MASK] tokens on MLMs. Analytical studies show that masking tokens can introduce the corrupted semantics problem, wherein the corrupted context may convey multiple, ambiguous meanings. This problem is also a key factor affecting the performance of MLMs on downstream tasks. Based"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.13397","kind":"arxiv","version":5},"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/2501.13397/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":"2501.13397","created_at":"2026-07-05T11:17:39.968705+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.13397v5","created_at":"2026-07-05T11:17:39.968705+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.13397","created_at":"2026-07-05T11:17:39.968705+00:00"},{"alias_kind":"pith_short_12","alias_value":"D3PFMI63QW7L","created_at":"2026-07-05T11:17:39.968705+00:00"},{"alias_kind":"pith_short_16","alias_value":"D3PFMI63QW7LYDSF","created_at":"2026-07-05T11:17:39.968705+00:00"},{"alias_kind":"pith_short_8","alias_value":"D3PFMI63","created_at":"2026-07-05T11:17:39.968705+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00052","citing_title":"AGE: Adaptive-masking for Graph Embedding in Graph Retrieval-Augmented Generation","ref_index":90,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN","json":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN.json","graph_json":"https://pith.science/api/pith-number/D3PFMI63QW7LYDSFENT2OKXKFN/graph.json","events_json":"https://pith.science/api/pith-number/D3PFMI63QW7LYDSFENT2OKXKFN/events.json","paper":"https://pith.science/paper/D3PFMI63"},"agent_actions":{"view_html":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN","download_json":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN.json","view_paper":"https://pith.science/paper/D3PFMI63","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.13397&json=true","fetch_graph":"https://pith.science/api/pith-number/D3PFMI63QW7LYDSFENT2OKXKFN/graph.json","fetch_events":"https://pith.science/api/pith-number/D3PFMI63QW7LYDSFENT2OKXKFN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN/action/storage_attestation","attest_author":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN/action/author_attestation","sign_citation":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN/action/citation_signature","submit_replication":"https://pith.science/pith/D3PFMI63QW7LYDSFENT2OKXKFN/action/replication_record"}},"created_at":"2026-07-05T11:17:39.968705+00:00","updated_at":"2026-07-05T11:17:39.968705+00:00"}