{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:ZB5ODL5XE63BOK7QBW4BFM2MJ5","short_pith_number":"pith:ZB5ODL5X","schema_version":"1.0","canonical_sha256":"c87ae1afb727b6172bf00db812b34c4f679b07b59d750fc52fcddc3292dd045e","source":{"kind":"arxiv","id":"2010.02329","version":4},"attestation_state":"computed","paper":{"title":"InfoBERT: Improving Robustness of Language Models from An Information Theoretic Perspective","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bo Li, Boxin Wang, Jingjing Liu, Ruoxi Jia, Shuohang Wang, Yu Cheng, Zhe Gan","submitted_at":"2020-10-05T20:49:26Z","abstract_excerpt":"Large-scale language models such as BERT have achieved state-of-the-art performance across a wide range of NLP tasks. Recent studies, however, show that such BERT-based models are vulnerable facing the threats of textual adversarial attacks. We aim to address this problem from an information-theoretic perspective, and propose InfoBERT, a novel learning framework for robust fine-tuning of pre-trained language models. InfoBERT contains two mutual-information-based regularizers for model training: (i) an Information Bottleneck regularizer, which suppresses noisy mutual information between the inp"},"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":"2010.02329","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-10-05T20:49:26Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"bdb1cabfd18a2fe345c292bfc49d7c38eba376dd4f256f78a576282baf45772f","abstract_canon_sha256":"0e36f72e8c9b97152c7cea87fde605b9a4f8e68db3015fb57a6eb9e0c1d9bf47"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:24:51.745377Z","signature_b64":"FsNxkFBn35H5NC+4TQaVOLkJn7ZJyif+0O4mZ0ayIEOkWelLyGn817WOh3Cx5DjN3T2KY4PWMbBFk+RCS6njDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c87ae1afb727b6172bf00db812b34c4f679b07b59d750fc52fcddc3292dd045e","last_reissued_at":"2026-07-05T02:24:51.744954Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:24:51.744954Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"InfoBERT: Improving Robustness of Language Models from An Information Theoretic Perspective","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bo Li, Boxin Wang, Jingjing Liu, Ruoxi Jia, Shuohang Wang, Yu Cheng, Zhe Gan","submitted_at":"2020-10-05T20:49:26Z","abstract_excerpt":"Large-scale language models such as BERT have achieved state-of-the-art performance across a wide range of NLP tasks. Recent studies, however, show that such BERT-based models are vulnerable facing the threats of textual adversarial attacks. We aim to address this problem from an information-theoretic perspective, and propose InfoBERT, a novel learning framework for robust fine-tuning of pre-trained language models. InfoBERT contains two mutual-information-based regularizers for model training: (i) an Information Bottleneck regularizer, which suppresses noisy mutual information between the inp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.02329","kind":"arxiv","version":4},"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/2010.02329/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":"2010.02329","created_at":"2026-07-05T02:24:51.745006+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.02329v4","created_at":"2026-07-05T02:24:51.745006+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.02329","created_at":"2026-07-05T02:24:51.745006+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZB5ODL5XE63B","created_at":"2026-07-05T02:24:51.745006+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZB5ODL5XE63BOK7Q","created_at":"2026-07-05T02:24:51.745006+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZB5ODL5X","created_at":"2026-07-05T02:24:51.745006+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2201.11990","citing_title":"Using DeepSpeed and Megatron to Train Megatron-Turing NLG 530B, A Large-Scale Generative Language Model","ref_index":68,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5","json":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5.json","graph_json":"https://pith.science/api/pith-number/ZB5ODL5XE63BOK7QBW4BFM2MJ5/graph.json","events_json":"https://pith.science/api/pith-number/ZB5ODL5XE63BOK7QBW4BFM2MJ5/events.json","paper":"https://pith.science/paper/ZB5ODL5X"},"agent_actions":{"view_html":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5","download_json":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5.json","view_paper":"https://pith.science/paper/ZB5ODL5X","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.02329&json=true","fetch_graph":"https://pith.science/api/pith-number/ZB5ODL5XE63BOK7QBW4BFM2MJ5/graph.json","fetch_events":"https://pith.science/api/pith-number/ZB5ODL5XE63BOK7QBW4BFM2MJ5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5/action/storage_attestation","attest_author":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5/action/author_attestation","sign_citation":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5/action/citation_signature","submit_replication":"https://pith.science/pith/ZB5ODL5XE63BOK7QBW4BFM2MJ5/action/replication_record"}},"created_at":"2026-07-05T02:24:51.745006+00:00","updated_at":"2026-07-05T02:24:51.745006+00:00"}