{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:OCZA23FAG6DVVEPHF4DQ6LQMMB","short_pith_number":"pith:OCZA23FA","schema_version":"1.0","canonical_sha256":"70b20d6ca037875a91e72f070f2e0c6053fe4f7b930af69f6f05ef5043d4fae5","source":{"kind":"arxiv","id":"2607.12767","version":1},"attestation_state":"computed","paper":{"title":"Accuracy and Normalized Accuracy under Length Bias: Analysis, Guidelines, and a Bayesian Alternative","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Koen Oostermeijer","submitted_at":"2026-07-14T13:39:11Z","abstract_excerpt":"Multiple-choice benchmarks that rank candidate completions by conditional log-probability suffer from a length bias: because log-probabilities sum over tokens, longer answers tend to be penalized relative to shorter ones in practice. A common mitigation is to normalize scores by completion length, but we show empirically that this heuristic frequently over-corrects, introducing a bias toward longer answers instead. We first analyze these scoring rules, characterizing when standard and length-normalized accuracy are appropriate and how their length biases depend on the distribution of completio"},"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":"2607.12767","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-14T13:39:11Z","cross_cats_sorted":[],"title_canon_sha256":"b81b077b065274e147da98513cfb8f829b226a06d2915611f8a94865d0c2be5c","abstract_canon_sha256":"d85cd240cb15c3dc028c100f0798a278a6d301aba91c02abdfa8e637abb5e4af"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-15T01:22:16.769955Z","signature_b64":"xfieiBUkPqY70q7E7gzAe6wJMtiSV559g/h5/D8/TKwuJOuUjwaPNNQXnzfRSoxc7jS2K9nItw5whhHAb06ZAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"70b20d6ca037875a91e72f070f2e0c6053fe4f7b930af69f6f05ef5043d4fae5","last_reissued_at":"2026-07-15T01:22:16.769137Z","signature_status":"signed_v1","first_computed_at":"2026-07-15T01:22:16.769137Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accuracy and Normalized Accuracy under Length Bias: Analysis, Guidelines, and a Bayesian Alternative","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Koen Oostermeijer","submitted_at":"2026-07-14T13:39:11Z","abstract_excerpt":"Multiple-choice benchmarks that rank candidate completions by conditional log-probability suffer from a length bias: because log-probabilities sum over tokens, longer answers tend to be penalized relative to shorter ones in practice. A common mitigation is to normalize scores by completion length, but we show empirically that this heuristic frequently over-corrects, introducing a bias toward longer answers instead. We first analyze these scoring rules, characterizing when standard and length-normalized accuracy are appropriate and how their length biases depend on the distribution of completio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.12767","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/2607.12767/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":"2607.12767","created_at":"2026-07-15T01:22:16.769563+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.12767v1","created_at":"2026-07-15T01:22:16.769563+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.12767","created_at":"2026-07-15T01:22:16.769563+00:00"},{"alias_kind":"pith_short_12","alias_value":"OCZA23FAG6DV","created_at":"2026-07-15T01:22:16.769563+00:00"},{"alias_kind":"pith_short_16","alias_value":"OCZA23FAG6DVVEPH","created_at":"2026-07-15T01:22:16.769563+00:00"},{"alias_kind":"pith_short_8","alias_value":"OCZA23FA","created_at":"2026-07-15T01:22:16.769563+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/OCZA23FAG6DVVEPHF4DQ6LQMMB","json":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB.json","graph_json":"https://pith.science/api/pith-number/OCZA23FAG6DVVEPHF4DQ6LQMMB/graph.json","events_json":"https://pith.science/api/pith-number/OCZA23FAG6DVVEPHF4DQ6LQMMB/events.json","paper":"https://pith.science/paper/OCZA23FA"},"agent_actions":{"view_html":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB","download_json":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB.json","view_paper":"https://pith.science/paper/OCZA23FA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.12767&json=true","fetch_graph":"https://pith.science/api/pith-number/OCZA23FAG6DVVEPHF4DQ6LQMMB/graph.json","fetch_events":"https://pith.science/api/pith-number/OCZA23FAG6DVVEPHF4DQ6LQMMB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB/action/storage_attestation","attest_author":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB/action/author_attestation","sign_citation":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB/action/citation_signature","submit_replication":"https://pith.science/pith/OCZA23FAG6DVVEPHF4DQ6LQMMB/action/replication_record"}},"created_at":"2026-07-15T01:22:16.769563+00:00","updated_at":"2026-07-15T01:22:16.769563+00:00"}