{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2021:4HQNHZBTVYWCFKVCWNDF5ZDEEG","short_pith_number":"pith:4HQNHZBT","canonical_record":{"source":{"id":"2105.14553","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-05-30T14:24:53Z","cross_cats_sorted":[],"title_canon_sha256":"4b211714c0e7e7cd880a460430046a9e2b45e87ebd570292c32849380567be22","abstract_canon_sha256":"b398222a4900a2f30a45b4a32351b637a579f37585fe6c5235daf944b5afbc84"},"schema_version":"1.0"},"canonical_sha256":"e1e0d3e433ae2c22aaa2b3465ee46421a837c207177b203bd07cd3367b3a4a3c","source":{"kind":"arxiv","id":"2105.14553","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2105.14553","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"arxiv_version","alias_value":"2105.14553v1","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2105.14553","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"pith_short_12","alias_value":"4HQNHZBTVYWC","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"pith_short_16","alias_value":"4HQNHZBTVYWCFKVC","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"pith_short_8","alias_value":"4HQNHZBT","created_at":"2026-07-05T02:44:38Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2021:4HQNHZBTVYWCFKVCWNDF5ZDEEG","target":"record","payload":{"canonical_record":{"source":{"id":"2105.14553","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-05-30T14:24:53Z","cross_cats_sorted":[],"title_canon_sha256":"4b211714c0e7e7cd880a460430046a9e2b45e87ebd570292c32849380567be22","abstract_canon_sha256":"b398222a4900a2f30a45b4a32351b637a579f37585fe6c5235daf944b5afbc84"},"schema_version":"1.0"},"canonical_sha256":"e1e0d3e433ae2c22aaa2b3465ee46421a837c207177b203bd07cd3367b3a4a3c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:44:38.865944Z","signature_b64":"MhVyy+xQdKpn0CnPRNyoasZWL8kc5aEkO3e4XCmOnkhFa8xf51bRdNR59+cyUTRnvRwuaVOHbZ2Kojw6GU+5DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e1e0d3e433ae2c22aaa2b3465ee46421a837c207177b203bd07cd3367b3a4a3c","last_reissued_at":"2026-07-05T02:44:38.865584Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:44:38.865584Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2105.14553","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T02:44:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3zeaT6aSv4l3J7AQu+fXtDbKWSXBCe3jmhudzzSxIOwyNMbtnPFFAR9//aDj59vP+BLsr+eHccXhIVDZZrEgAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T22:00:16.033875Z"},"content_sha256":"883af83ac8481e7b7b35d850469e7832d85c47ab8cfd4237ce4f93d8bd534c62","schema_version":"1.0","event_id":"sha256:883af83ac8481e7b7b35d850469e7832d85c47ab8cfd4237ce4f93d8bd534c62"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2021:4HQNHZBTVYWCFKVCWNDF5ZDEEG","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Defending Pre-trained Language Models from Adversarial Word Substitutions Without Performance Sacrifice","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hai Zhao, Jiayi Wang, Rongzhou Bao","submitted_at":"2021-05-30T14:24:53Z","abstract_excerpt":"Pre-trained contextualized language models (PrLMs) have led to strong performance gains in downstream natural language understanding tasks. However, PrLMs can still be easily fooled by adversarial word substitution, which is one of the most challenging textual adversarial attack methods. Existing defence approaches suffer from notable performance loss and complexities. Thus, this paper presents a compact and performance-preserved framework, Anomaly Detection with Frequency-Aware Randomization (ADFAR). In detail, we design an auxiliary anomaly detection classifier and adopt a multi-task learnin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2105.14553","kind":"arxiv","version":1},"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/2105.14553/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T02:44:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"EgH92cIG+TvkF8c9CmiK0A+Zcf0jZRXp7f5g+iWoDymTVUyfb5pTJw5kc9xNGitcByOkCkZTdjzmyqb7htROAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T22:00:16.034692Z"},"content_sha256":"d27595ce55dfeadfcbaf79747feb3ceaba7b1c31c41517041d7e83ac2dda08a0","schema_version":"1.0","event_id":"sha256:d27595ce55dfeadfcbaf79747feb3ceaba7b1c31c41517041d7e83ac2dda08a0"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG/bundle.json","state_url":"https://pith.science/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-11T22:00:16Z","links":{"resolver":"https://pith.science/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG","bundle":"https://pith.science/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG/bundle.json","state":"https://pith.science/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG/state.json","well_known_bundle":"https://pith.science/.well-known/pith/4HQNHZBTVYWCFKVCWNDF5ZDEEG/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2021:4HQNHZBTVYWCFKVCWNDF5ZDEEG","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"b398222a4900a2f30a45b4a32351b637a579f37585fe6c5235daf944b5afbc84","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-05-30T14:24:53Z","title_canon_sha256":"4b211714c0e7e7cd880a460430046a9e2b45e87ebd570292c32849380567be22"},"schema_version":"1.0","source":{"id":"2105.14553","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2105.14553","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"arxiv_version","alias_value":"2105.14553v1","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2105.14553","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"pith_short_12","alias_value":"4HQNHZBTVYWC","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"pith_short_16","alias_value":"4HQNHZBTVYWCFKVC","created_at":"2026-07-05T02:44:38Z"},{"alias_kind":"pith_short_8","alias_value":"4HQNHZBT","created_at":"2026-07-05T02:44:38Z"}],"graph_snapshots":[{"event_id":"sha256:d27595ce55dfeadfcbaf79747feb3ceaba7b1c31c41517041d7e83ac2dda08a0","target":"graph","created_at":"2026-07-05T02:44:38Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2105.14553/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Pre-trained contextualized language models (PrLMs) have led to strong performance gains in downstream natural language understanding tasks. However, PrLMs can still be easily fooled by adversarial word substitution, which is one of the most challenging textual adversarial attack methods. Existing defence approaches suffer from notable performance loss and complexities. Thus, this paper presents a compact and performance-preserved framework, Anomaly Detection with Frequency-Aware Randomization (ADFAR). In detail, we design an auxiliary anomaly detection classifier and adopt a multi-task learnin","authors_text":"Hai Zhao, Jiayi Wang, Rongzhou Bao","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-05-30T14:24:53Z","title":"Defending Pre-trained Language Models from Adversarial Word Substitutions Without Performance Sacrifice"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2105.14553","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:883af83ac8481e7b7b35d850469e7832d85c47ab8cfd4237ce4f93d8bd534c62","target":"record","created_at":"2026-07-05T02:44:38Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"b398222a4900a2f30a45b4a32351b637a579f37585fe6c5235daf944b5afbc84","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2021-05-30T14:24:53Z","title_canon_sha256":"4b211714c0e7e7cd880a460430046a9e2b45e87ebd570292c32849380567be22"},"schema_version":"1.0","source":{"id":"2105.14553","kind":"arxiv","version":1}},"canonical_sha256":"e1e0d3e433ae2c22aaa2b3465ee46421a837c207177b203bd07cd3367b3a4a3c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e1e0d3e433ae2c22aaa2b3465ee46421a837c207177b203bd07cd3367b3a4a3c","first_computed_at":"2026-07-05T02:44:38.865584Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T02:44:38.865584Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"MhVyy+xQdKpn0CnPRNyoasZWL8kc5aEkO3e4XCmOnkhFa8xf51bRdNR59+cyUTRnvRwuaVOHbZ2Kojw6GU+5DQ==","signature_status":"signed_v1","signed_at":"2026-07-05T02:44:38.865944Z","signed_message":"canonical_sha256_bytes"},"source_id":"2105.14553","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:883af83ac8481e7b7b35d850469e7832d85c47ab8cfd4237ce4f93d8bd534c62","sha256:d27595ce55dfeadfcbaf79747feb3ceaba7b1c31c41517041d7e83ac2dda08a0"],"state_sha256":"2155d243c6672aa656a517711b17360d7923a5fae2b87e798b61fb218bfff31a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"gJlfRLjN8LcS6R1BQBPwBAp1wB39WatIXhYeJmHCr1L1BymaxUgFze/bG8VPgPiJPCaGXxj+eQCkHcIRPNB8Cg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-11T22:00:16.041507Z","bundle_sha256":"8afd257acb0505f87c386965434accc84b44a8006878ee6583ad45ca9218ad70"}}