{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:QSQQWO3POZTNCAV5UCLSV3XRPT","short_pith_number":"pith:QSQQWO3P","schema_version":"1.0","canonical_sha256":"84a10b3b6f7666d102bda0972aeef17cd8fe01291a37cf0b0b4f575203e3ade0","source":{"kind":"arxiv","id":"2106.01024","version":1},"attestation_state":"computed","paper":{"title":"Why Machine Reading Comprehension Models Learn Shortcuts?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chen Zhang, Dongyan Zhao, Quzhe Huang, Yansong Feng, Yuxuan Lai","submitted_at":"2021-06-02T08:43:12Z","abstract_excerpt":"Recent studies report that many machine reading comprehension (MRC) models can perform closely to or even better than humans on benchmark datasets. However, existing works indicate that many MRC models may learn shortcuts to outwit these benchmarks, but the performance is unsatisfactory in real-world applications. In this work, we attempt to explore, instead of the expected comprehension skills, why these models learn the shortcuts. Based on the observation that a large portion of questions in current datasets have shortcut solutions, we argue that larger proportion of shortcut questions in tr"},"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":"2106.01024","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-06-02T08:43:12Z","cross_cats_sorted":[],"title_canon_sha256":"8b183a2f5035b0d7be0fa243c3b00a540ab7db66d43b391c1f3c43f2a7df5523","abstract_canon_sha256":"538a5b2cc8c23feac40be3f71a5ebc6597bd4b02b0111f15eea2b8bd08d78718"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:45:43.010964Z","signature_b64":"7JMNJ5pT9oUhpMI3q1Q7Ae+MiOOnb7USmbfL0MrIFXTcmoYRKmDgbL5cd/r9IYLeBVyArfvxbVwCZe2wfNVYBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"84a10b3b6f7666d102bda0972aeef17cd8fe01291a37cf0b0b4f575203e3ade0","last_reissued_at":"2026-07-05T02:45:43.010536Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:45:43.010536Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why Machine Reading Comprehension Models Learn Shortcuts?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chen Zhang, Dongyan Zhao, Quzhe Huang, Yansong Feng, Yuxuan Lai","submitted_at":"2021-06-02T08:43:12Z","abstract_excerpt":"Recent studies report that many machine reading comprehension (MRC) models can perform closely to or even better than humans on benchmark datasets. However, existing works indicate that many MRC models may learn shortcuts to outwit these benchmarks, but the performance is unsatisfactory in real-world applications. In this work, we attempt to explore, instead of the expected comprehension skills, why these models learn the shortcuts. Based on the observation that a large portion of questions in current datasets have shortcut solutions, we argue that larger proportion of shortcut questions in tr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.01024","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/2106.01024/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":"2106.01024","created_at":"2026-07-05T02:45:43.010595+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.01024v1","created_at":"2026-07-05T02:45:43.010595+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.01024","created_at":"2026-07-05T02:45:43.010595+00:00"},{"alias_kind":"pith_short_12","alias_value":"QSQQWO3POZTN","created_at":"2026-07-05T02:45:43.010595+00:00"},{"alias_kind":"pith_short_16","alias_value":"QSQQWO3POZTNCAV5","created_at":"2026-07-05T02:45:43.010595+00:00"},{"alias_kind":"pith_short_8","alias_value":"QSQQWO3P","created_at":"2026-07-05T02:45:43.010595+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT","json":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT.json","graph_json":"https://pith.science/api/pith-number/QSQQWO3POZTNCAV5UCLSV3XRPT/graph.json","events_json":"https://pith.science/api/pith-number/QSQQWO3POZTNCAV5UCLSV3XRPT/events.json","paper":"https://pith.science/paper/QSQQWO3P"},"agent_actions":{"view_html":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT","download_json":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT.json","view_paper":"https://pith.science/paper/QSQQWO3P","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.01024&json=true","fetch_graph":"https://pith.science/api/pith-number/QSQQWO3POZTNCAV5UCLSV3XRPT/graph.json","fetch_events":"https://pith.science/api/pith-number/QSQQWO3POZTNCAV5UCLSV3XRPT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT/action/storage_attestation","attest_author":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT/action/author_attestation","sign_citation":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT/action/citation_signature","submit_replication":"https://pith.science/pith/QSQQWO3POZTNCAV5UCLSV3XRPT/action/replication_record"}},"created_at":"2026-07-05T02:45:43.010595+00:00","updated_at":"2026-07-05T02:45:43.010595+00:00"}