{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2020:SSBXNQSXGHJDGPJPTDMM22YA4T","short_pith_number":"pith:SSBXNQSX","canonical_record":{"source":{"id":"2004.08994","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-20T00:07:18Z","cross_cats_sorted":[],"title_canon_sha256":"9a16e9158ca2dc16b0c7dc4bf59c4f57b9180af9b2cb058cb7fece8141c11d4a","abstract_canon_sha256":"c96d0c37577ea78fc1886f0f67266a4341d81b757cc4a4a3ee08949dbba09338"},"schema_version":"1.0"},"canonical_sha256":"948376c25731d2333d2f98d8cd6b00e4c362571efef4bb2ca49dddb4b97a7fe7","source":{"kind":"arxiv","id":"2004.08994","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2004.08994","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"arxiv_version","alias_value":"2004.08994v2","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.08994","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"pith_short_12","alias_value":"SSBXNQSXGHJD","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"pith_short_16","alias_value":"SSBXNQSXGHJDGPJP","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"pith_short_8","alias_value":"SSBXNQSX","created_at":"2026-07-05T00:59:24Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2020:SSBXNQSXGHJDGPJPTDMM22YA4T","target":"record","payload":{"canonical_record":{"source":{"id":"2004.08994","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-20T00:07:18Z","cross_cats_sorted":[],"title_canon_sha256":"9a16e9158ca2dc16b0c7dc4bf59c4f57b9180af9b2cb058cb7fece8141c11d4a","abstract_canon_sha256":"c96d0c37577ea78fc1886f0f67266a4341d81b757cc4a4a3ee08949dbba09338"},"schema_version":"1.0"},"canonical_sha256":"948376c25731d2333d2f98d8cd6b00e4c362571efef4bb2ca49dddb4b97a7fe7","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:59:24.971758Z","signature_b64":"RJEPm0akHcuuj3K0PccJ8ZqmaXIIi933lES9d9/bpUEl9myL+Jf2qKSSA5z/ou/ZoyZIXHQjEOEuHtS7PdbfCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"948376c25731d2333d2f98d8cd6b00e4c362571efef4bb2ca49dddb4b97a7fe7","last_reissued_at":"2026-07-05T00:59:24.971371Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:59:24.971371Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2004.08994","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T00:59:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mCpzLszeQxa5WjkOPdcz2mklFbdGHYinVZos8/hNu2Y79arGbwwYOw4G3XAouo1twiaQdOBrP1WMCN/QvTgGCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T14:39:54.752563Z"},"content_sha256":"5bfb914b171c8641ba74534f8ef406cfd89018eb73aac6703a0c38aba06174b8","schema_version":"1.0","event_id":"sha256:5bfb914b171c8641ba74534f8ef406cfd89018eb73aac6703a0c38aba06174b8"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2020:SSBXNQSXGHJDGPJPTDMM22YA4T","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Adversarial Training for Large Neural Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hao Cheng, Hoifung Poon, Jianfeng Gao, Pengcheng He, Weizhu Chen, XiaoDong Liu, Yu Wang","submitted_at":"2020-04-20T00:07:18Z","abstract_excerpt":"Generalization and robustness are both key desiderata for designing machine learning methods. Adversarial training can enhance robustness, but past work often finds it hurts generalization. In natural language processing (NLP), pre-training large neural language models such as BERT have demonstrated impressive gain in generalization for a variety of tasks, with further improvement from adversarial fine-tuning. However, these models are still vulnerable to adversarial attacks. In this paper, we show that adversarial pre-training can improve both generalization and robustness. We propose a gener"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.08994","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/2004.08994/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":"2026-07-05T00:59:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AubWVlp3nI8OfaoAA8f2pNFjbvxUFN34EVc9mKkelApXXr7kaYM4IrRqijFqa3CN5cf1ZE+9aUMteqFnT8eqAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T14:39:54.752939Z"},"content_sha256":"f48b5bd56617b358ee231199085015f53ad843c44e9521bf54cfb32c17523e6d","schema_version":"1.0","event_id":"sha256:f48b5bd56617b358ee231199085015f53ad843c44e9521bf54cfb32c17523e6d"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/SSBXNQSXGHJDGPJPTDMM22YA4T/bundle.json","state_url":"https://pith.science/pith/SSBXNQSXGHJDGPJPTDMM22YA4T/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/SSBXNQSXGHJDGPJPTDMM22YA4T/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":"2026-07-07T14:39:54Z","links":{"resolver":"https://pith.science/pith/SSBXNQSXGHJDGPJPTDMM22YA4T","bundle":"https://pith.science/pith/SSBXNQSXGHJDGPJPTDMM22YA4T/bundle.json","state":"https://pith.science/pith/SSBXNQSXGHJDGPJPTDMM22YA4T/state.json","well_known_bundle":"https://pith.science/.well-known/pith/SSBXNQSXGHJDGPJPTDMM22YA4T/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2020:SSBXNQSXGHJDGPJPTDMM22YA4T","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":"c96d0c37577ea78fc1886f0f67266a4341d81b757cc4a4a3ee08949dbba09338","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-20T00:07:18Z","title_canon_sha256":"9a16e9158ca2dc16b0c7dc4bf59c4f57b9180af9b2cb058cb7fece8141c11d4a"},"schema_version":"1.0","source":{"id":"2004.08994","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2004.08994","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"arxiv_version","alias_value":"2004.08994v2","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.08994","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"pith_short_12","alias_value":"SSBXNQSXGHJD","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"pith_short_16","alias_value":"SSBXNQSXGHJDGPJP","created_at":"2026-07-05T00:59:24Z"},{"alias_kind":"pith_short_8","alias_value":"SSBXNQSX","created_at":"2026-07-05T00:59:24Z"}],"graph_snapshots":[{"event_id":"sha256:f48b5bd56617b358ee231199085015f53ad843c44e9521bf54cfb32c17523e6d","target":"graph","created_at":"2026-07-05T00:59:24Z","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/2004.08994/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Generalization and robustness are both key desiderata for designing machine learning methods. Adversarial training can enhance robustness, but past work often finds it hurts generalization. In natural language processing (NLP), pre-training large neural language models such as BERT have demonstrated impressive gain in generalization for a variety of tasks, with further improvement from adversarial fine-tuning. However, these models are still vulnerable to adversarial attacks. In this paper, we show that adversarial pre-training can improve both generalization and robustness. We propose a gener","authors_text":"Hao Cheng, Hoifung Poon, Jianfeng Gao, Pengcheng He, Weizhu Chen, XiaoDong Liu, Yu Wang","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-20T00:07:18Z","title":"Adversarial Training for Large Neural Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.08994","kind":"arxiv","version":2},"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":"sha256:5bfb914b171c8641ba74534f8ef406cfd89018eb73aac6703a0c38aba06174b8","target":"record","created_at":"2026-07-05T00:59:24Z","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":"c96d0c37577ea78fc1886f0f67266a4341d81b757cc4a4a3ee08949dbba09338","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-20T00:07:18Z","title_canon_sha256":"9a16e9158ca2dc16b0c7dc4bf59c4f57b9180af9b2cb058cb7fece8141c11d4a"},"schema_version":"1.0","source":{"id":"2004.08994","kind":"arxiv","version":2}},"canonical_sha256":"948376c25731d2333d2f98d8cd6b00e4c362571efef4bb2ca49dddb4b97a7fe7","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"948376c25731d2333d2f98d8cd6b00e4c362571efef4bb2ca49dddb4b97a7fe7","first_computed_at":"2026-07-05T00:59:24.971371Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T00:59:24.971371Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"RJEPm0akHcuuj3K0PccJ8ZqmaXIIi933lES9d9/bpUEl9myL+Jf2qKSSA5z/ou/ZoyZIXHQjEOEuHtS7PdbfCw==","signature_status":"signed_v1","signed_at":"2026-07-05T00:59:24.971758Z","signed_message":"canonical_sha256_bytes"},"source_id":"2004.08994","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5bfb914b171c8641ba74534f8ef406cfd89018eb73aac6703a0c38aba06174b8","sha256:f48b5bd56617b358ee231199085015f53ad843c44e9521bf54cfb32c17523e6d"],"state_sha256":"0b95f08fa74c0b74574f77698b9699c0d51c98a1d080f1f4838458d31950fdf0"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WgU/Gp5hxCC+RHvPdBGRWAC1BKTtlQ+jeUzZNxX3zBZvO1HWHB+GVTJN35UoUd80gtCskobH4xLV4njCK+EJBA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T14:39:54.754929Z","bundle_sha256":"6c8608fcf42857ca82eb7e498980fbe3b02fb729096fde0fdcd4a96548fb1753"}}