{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:UXH5OU23URAK47NS4YEKZKUV3M","short_pith_number":"pith:UXH5OU23","schema_version":"1.0","canonical_sha256":"a5cfd7535ba440ae7db2e608acaa95db1fb66e471a18e45da05a7a4b192abdf1","source":{"kind":"arxiv","id":"2010.04736","version":1},"attestation_state":"computed","paper":{"title":"Evaluating and Characterizing Human Rationales","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.HC","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anirudh Rathore, Chenhao Tan, Samuel Carton","submitted_at":"2020-10-09T18:00:04Z","abstract_excerpt":"Two main approaches for evaluating the quality of machine-generated rationales are: 1) using human rationales as a gold standard; and 2) automated metrics based on how rationales affect model behavior. An open question, however, is how human rationales fare with these automatic metrics. Analyzing a variety of datasets and models, we find that human rationales do not necessarily perform well on these metrics. To unpack this finding, we propose improved metrics to account for model-dependent baseline performance. We then propose two methods to further characterize rationale quality, one based on"},"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":"2010.04736","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-10-09T18:00:04Z","cross_cats_sorted":["cs.AI","cs.CY","cs.HC","cs.LG"],"title_canon_sha256":"14970146174806336930a7bb9ff4953b69065b1e851f4670be80aae1cc31df8d","abstract_canon_sha256":"26a6b4d50d74ded879b14f18e9c5f6c55f1c50e312ea2c222a545dd937ca729d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:41:58.655499Z","signature_b64":"fJ+hRqnsrC8brbpvaYBHf5qxdIh0HlayD3DeQ+OnYcOtqEX+QkEcXTAcXXmfBSvK8/WOunNpgQhrIJ43jOCPBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a5cfd7535ba440ae7db2e608acaa95db1fb66e471a18e45da05a7a4b192abdf1","last_reissued_at":"2026-07-05T01:41:58.655005Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:41:58.655005Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evaluating and Characterizing Human Rationales","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CY","cs.HC","cs.LG"],"primary_cat":"cs.CL","authors_text":"Anirudh Rathore, Chenhao Tan, Samuel Carton","submitted_at":"2020-10-09T18:00:04Z","abstract_excerpt":"Two main approaches for evaluating the quality of machine-generated rationales are: 1) using human rationales as a gold standard; and 2) automated metrics based on how rationales affect model behavior. An open question, however, is how human rationales fare with these automatic metrics. Analyzing a variety of datasets and models, we find that human rationales do not necessarily perform well on these metrics. To unpack this finding, we propose improved metrics to account for model-dependent baseline performance. We then propose two methods to further characterize rationale quality, one based on"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.04736","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/2010.04736/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":"2010.04736","created_at":"2026-07-05T01:41:58.655066+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.04736v1","created_at":"2026-07-05T01:41:58.655066+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.04736","created_at":"2026-07-05T01:41:58.655066+00:00"},{"alias_kind":"pith_short_12","alias_value":"UXH5OU23URAK","created_at":"2026-07-05T01:41:58.655066+00:00"},{"alias_kind":"pith_short_16","alias_value":"UXH5OU23URAK47NS","created_at":"2026-07-05T01:41:58.655066+00:00"},{"alias_kind":"pith_short_8","alias_value":"UXH5OU23","created_at":"2026-07-05T01:41:58.655066+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.21302","citing_title":"Can human clinical rationales improve the performance and explainability of clinical text classification models?","ref_index":3,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M","json":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M.json","graph_json":"https://pith.science/api/pith-number/UXH5OU23URAK47NS4YEKZKUV3M/graph.json","events_json":"https://pith.science/api/pith-number/UXH5OU23URAK47NS4YEKZKUV3M/events.json","paper":"https://pith.science/paper/UXH5OU23"},"agent_actions":{"view_html":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M","download_json":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M.json","view_paper":"https://pith.science/paper/UXH5OU23","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.04736&json=true","fetch_graph":"https://pith.science/api/pith-number/UXH5OU23URAK47NS4YEKZKUV3M/graph.json","fetch_events":"https://pith.science/api/pith-number/UXH5OU23URAK47NS4YEKZKUV3M/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M/action/storage_attestation","attest_author":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M/action/author_attestation","sign_citation":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M/action/citation_signature","submit_replication":"https://pith.science/pith/UXH5OU23URAK47NS4YEKZKUV3M/action/replication_record"}},"created_at":"2026-07-05T01:41:58.655066+00:00","updated_at":"2026-07-05T01:41:58.655066+00:00"}