{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JQNSD25VQTFNQJRJSHZAXYTUEJ","short_pith_number":"pith:JQNSD25V","schema_version":"1.0","canonical_sha256":"4c1b21ebb584cad8262991f20be2742258e0cda3e89900335ab3a7aa6703bbf4","source":{"kind":"arxiv","id":"2504.11972","version":3},"attestation_state":"computed","paper":{"title":"Reassessing Extractive QA Datasets at Scale: LLM-as-a-Judge and In-Depth Analyses","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Akiko Aizawa, Florian Boudin, Jiahao Huang, Xanh Ho","submitted_at":"2025-04-16T11:08:46Z","abstract_excerpt":"Extractive QA tasks are commonly evaluated using Exact Match (EM) and F1-score, but these metrics often fail to reflect true model performance. Recent studies have proposed using large language models (LLMs) as judges (LLM-as-a-judge), yet they often lack comprehensive evaluation across datasets and overlook key factors such as sensitivity to answer types, prompt variations, and self-preference bias. In this work, we conduct a systematic study of LLM-as-a-judge across four extractive QA datasets and various prompt variations, assessing multiple LLM families in both answering and judging roles."},"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":"2504.11972","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-04-16T11:08:46Z","cross_cats_sorted":[],"title_canon_sha256":"495b477d1e1aa5f28d7ed018a570cc1c3dad38450dc1b07c7c83978873752b1f","abstract_canon_sha256":"2656d2a1fc52d5b6c98ca010718ebbcabbeb9e18b30ace7f66b23e3ee1c24639"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:02:15.883715Z","signature_b64":"yI4uVn9LNXUbJmoK1LeVAeqM7tGzUdH+PnGq93ljWtsHDIviDHycdCOCplIsmFErFowEy4w/3V5EKYtrT9M3Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4c1b21ebb584cad8262991f20be2742258e0cda3e89900335ab3a7aa6703bbf4","last_reissued_at":"2026-06-01T01:02:15.882626Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:02:15.882626Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reassessing Extractive QA Datasets at Scale: LLM-as-a-Judge and In-Depth Analyses","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Akiko Aizawa, Florian Boudin, Jiahao Huang, Xanh Ho","submitted_at":"2025-04-16T11:08:46Z","abstract_excerpt":"Extractive QA tasks are commonly evaluated using Exact Match (EM) and F1-score, but these metrics often fail to reflect true model performance. Recent studies have proposed using large language models (LLMs) as judges (LLM-as-a-judge), yet they often lack comprehensive evaluation across datasets and overlook key factors such as sensitivity to answer types, prompt variations, and self-preference bias. In this work, we conduct a systematic study of LLM-as-a-judge across four extractive QA datasets and various prompt variations, assessing multiple LLM families in both answering and judging roles."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.11972","kind":"arxiv","version":3},"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/2504.11972/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":"2504.11972","created_at":"2026-06-01T01:02:15.882779+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.11972v3","created_at":"2026-06-01T01:02:15.882779+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.11972","created_at":"2026-06-01T01:02:15.882779+00:00"},{"alias_kind":"pith_short_12","alias_value":"JQNSD25VQTFN","created_at":"2026-06-01T01:02:15.882779+00:00"},{"alias_kind":"pith_short_16","alias_value":"JQNSD25VQTFNQJRJ","created_at":"2026-06-01T01:02:15.882779+00:00"},{"alias_kind":"pith_short_8","alias_value":"JQNSD25V","created_at":"2026-06-01T01:02:15.882779+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.09791","citing_title":"Pioneer Agent: Continual Improvement of Small Language Models in Production","ref_index":36,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ","json":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ.json","graph_json":"https://pith.science/api/pith-number/JQNSD25VQTFNQJRJSHZAXYTUEJ/graph.json","events_json":"https://pith.science/api/pith-number/JQNSD25VQTFNQJRJSHZAXYTUEJ/events.json","paper":"https://pith.science/paper/JQNSD25V"},"agent_actions":{"view_html":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ","download_json":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ.json","view_paper":"https://pith.science/paper/JQNSD25V","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.11972&json=true","fetch_graph":"https://pith.science/api/pith-number/JQNSD25VQTFNQJRJSHZAXYTUEJ/graph.json","fetch_events":"https://pith.science/api/pith-number/JQNSD25VQTFNQJRJSHZAXYTUEJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ/action/storage_attestation","attest_author":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ/action/author_attestation","sign_citation":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ/action/citation_signature","submit_replication":"https://pith.science/pith/JQNSD25VQTFNQJRJSHZAXYTUEJ/action/replication_record"}},"created_at":"2026-06-01T01:02:15.882779+00:00","updated_at":"2026-06-01T01:02:15.882779+00:00"}