{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:HISKM452ODBNTYLG7CE7L3AD3Q","short_pith_number":"pith:HISKM452","schema_version":"1.0","canonical_sha256":"3a24a673ba70c2d9e166f889f5ec03dc3a095739d9e9f350a3fe91873f7905cb","source":{"kind":"arxiv","id":"2110.01518","version":1},"attestation_state":"computed","paper":{"title":"Generalization in NLI: Ways (Not) To Go Beyond Simple Heuristics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aleksandr Drozd, Anna Rogers, Prajjwal Bhargava","submitted_at":"2021-10-04T15:37:07Z","abstract_excerpt":"Much of recent progress in NLU was shown to be due to models' learning dataset-specific heuristics. We conduct a case study of generalization in NLI (from MNLI to the adversarially constructed HANS dataset) in a range of BERT-based architectures (adapters, Siamese Transformers, HEX debiasing), as well as with subsampling the data and increasing the model size. We report 2 successful and 3 unsuccessful strategies, all providing insights into how Transformer-based models learn to generalize."},"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":"2110.01518","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-10-04T15:37:07Z","cross_cats_sorted":[],"title_canon_sha256":"94237648edb49d5e9d97175e5d5855cfc411f37fa1b5bde367bf9178d7f2bf1b","abstract_canon_sha256":"5f5c53f05ace3542610574523f3bc767fc21e7c7b88906fa1156200c3d8c3e4a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:19:56.451090Z","signature_b64":"TX5lpiUcB9+eNS5BiNZg4w5WQ8ukCmqB++E7MgC90a12evyZTeAbpi9D6wstJ9E7bA3OCKvp6jNeUvKq7bsvCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3a24a673ba70c2d9e166f889f5ec03dc3a095739d9e9f350a3fe91873f7905cb","last_reissued_at":"2026-07-05T03:19:56.450695Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:19:56.450695Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Generalization in NLI: Ways (Not) To Go Beyond Simple Heuristics","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aleksandr Drozd, Anna Rogers, Prajjwal Bhargava","submitted_at":"2021-10-04T15:37:07Z","abstract_excerpt":"Much of recent progress in NLU was shown to be due to models' learning dataset-specific heuristics. We conduct a case study of generalization in NLI (from MNLI to the adversarially constructed HANS dataset) in a range of BERT-based architectures (adapters, Siamese Transformers, HEX debiasing), as well as with subsampling the data and increasing the model size. We report 2 successful and 3 unsuccessful strategies, all providing insights into how Transformer-based models learn to generalize."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2110.01518","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/2110.01518/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":"2110.01518","created_at":"2026-07-05T03:19:56.450752+00:00"},{"alias_kind":"arxiv_version","alias_value":"2110.01518v1","created_at":"2026-07-05T03:19:56.450752+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2110.01518","created_at":"2026-07-05T03:19:56.450752+00:00"},{"alias_kind":"pith_short_12","alias_value":"HISKM452ODBN","created_at":"2026-07-05T03:19:56.450752+00:00"},{"alias_kind":"pith_short_16","alias_value":"HISKM452ODBNTYLG","created_at":"2026-07-05T03:19:56.450752+00:00"},{"alias_kind":"pith_short_8","alias_value":"HISKM452","created_at":"2026-07-05T03:19:56.450752+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21240","citing_title":"DIPBox: A Multi-scale Testing Framework for Tracking Dataset Regeneration","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2509.23175","citing_title":"WARBERT: A Hierarchical BERT-based Model for Web API Recommendation","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04901","citing_title":"On the (In-)Security of the Shuffling Defense in the Transformer Secure Inference","ref_index":197,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q","json":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q.json","graph_json":"https://pith.science/api/pith-number/HISKM452ODBNTYLG7CE7L3AD3Q/graph.json","events_json":"https://pith.science/api/pith-number/HISKM452ODBNTYLG7CE7L3AD3Q/events.json","paper":"https://pith.science/paper/HISKM452"},"agent_actions":{"view_html":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q","download_json":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q.json","view_paper":"https://pith.science/paper/HISKM452","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2110.01518&json=true","fetch_graph":"https://pith.science/api/pith-number/HISKM452ODBNTYLG7CE7L3AD3Q/graph.json","fetch_events":"https://pith.science/api/pith-number/HISKM452ODBNTYLG7CE7L3AD3Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q/action/storage_attestation","attest_author":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q/action/author_attestation","sign_citation":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q/action/citation_signature","submit_replication":"https://pith.science/pith/HISKM452ODBNTYLG7CE7L3AD3Q/action/replication_record"}},"created_at":"2026-07-05T03:19:56.450752+00:00","updated_at":"2026-07-05T03:19:56.450752+00:00"}