{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:GBWTCE3MGKAVOM3KZV6CLBJKOR","short_pith_number":"pith:GBWTCE3M","schema_version":"1.0","canonical_sha256":"306d31136c328157336acd7c25852a745e4a94cceeefc1f84bc4710a242db2cb","source":{"kind":"arxiv","id":"2201.12091","version":8},"attestation_state":"computed","paper":{"title":"Linear Adversarial Concept Erasure","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Michael Twiton, Ryan Cotterell, Shauli Ravfogel, Yoav Goldberg","submitted_at":"2022-01-28T13:00:17Z","abstract_excerpt":"Modern neural models trained on textual data rely on pre-trained representations that emerge without direct supervision. As these representations are increasingly being used in real-world applications, the inability to \\emph{control} their content becomes an increasingly important problem. We formulate the problem of identifying and erasing a linear subspace that corresponds to a given concept, in order to prevent linear predictors from recovering the concept. We model this problem as a constrained, linear maximin game, and show that existing solutions are generally not optimal for this task. "},"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":"2201.12091","kind":"arxiv","version":8},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-01-28T13:00:17Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"4c4f9624cc6bf448d5000f1d353cd104d75c059de95372899c8243423aa4a4a8","abstract_canon_sha256":"4fa19b63f3ddf2c4caf9d378ce23cbc456084ea4a5c1b451565cc9135d3bf686"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:49:57.328155Z","signature_b64":"Q3pf6qMtqN43/z06PKRoYl1pu5HdADmuU/tc6DUejEX9orr94bGpiTRSQGN4nrzaF0iXJree9uLXHmkIQojtDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"306d31136c328157336acd7c25852a745e4a94cceeefc1f84bc4710a242db2cb","last_reissued_at":"2026-07-05T09:49:57.327604Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:49:57.327604Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Linear Adversarial Concept Erasure","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Michael Twiton, Ryan Cotterell, Shauli Ravfogel, Yoav Goldberg","submitted_at":"2022-01-28T13:00:17Z","abstract_excerpt":"Modern neural models trained on textual data rely on pre-trained representations that emerge without direct supervision. As these representations are increasingly being used in real-world applications, the inability to \\emph{control} their content becomes an increasingly important problem. We formulate the problem of identifying and erasing a linear subspace that corresponds to a given concept, in order to prevent linear predictors from recovering the concept. We model this problem as a constrained, linear maximin game, and show that existing solutions are generally not optimal for this task. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2201.12091","kind":"arxiv","version":8},"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/2201.12091/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":"2201.12091","created_at":"2026-07-05T09:49:57.327664+00:00"},{"alias_kind":"arxiv_version","alias_value":"2201.12091v8","created_at":"2026-07-05T09:49:57.327664+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2201.12091","created_at":"2026-07-05T09:49:57.327664+00:00"},{"alias_kind":"pith_short_12","alias_value":"GBWTCE3MGKAV","created_at":"2026-07-05T09:49:57.327664+00:00"},{"alias_kind":"pith_short_16","alias_value":"GBWTCE3MGKAVOM3K","created_at":"2026-07-05T09:49:57.327664+00:00"},{"alias_kind":"pith_short_8","alias_value":"GBWTCE3M","created_at":"2026-07-05T09:49:57.327664+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.07731","citing_title":"Energy-Guided Decoding for Object Hallucination Mitigation","ref_index":43,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR","json":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR.json","graph_json":"https://pith.science/api/pith-number/GBWTCE3MGKAVOM3KZV6CLBJKOR/graph.json","events_json":"https://pith.science/api/pith-number/GBWTCE3MGKAVOM3KZV6CLBJKOR/events.json","paper":"https://pith.science/paper/GBWTCE3M"},"agent_actions":{"view_html":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR","download_json":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR.json","view_paper":"https://pith.science/paper/GBWTCE3M","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2201.12091&json=true","fetch_graph":"https://pith.science/api/pith-number/GBWTCE3MGKAVOM3KZV6CLBJKOR/graph.json","fetch_events":"https://pith.science/api/pith-number/GBWTCE3MGKAVOM3KZV6CLBJKOR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR/action/storage_attestation","attest_author":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR/action/author_attestation","sign_citation":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR/action/citation_signature","submit_replication":"https://pith.science/pith/GBWTCE3MGKAVOM3KZV6CLBJKOR/action/replication_record"}},"created_at":"2026-07-05T09:49:57.327664+00:00","updated_at":"2026-07-05T09:49:57.327664+00:00"}