{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:FPEHY2L64U6N532CNFCM4BYWUC","short_pith_number":"pith:FPEHY2L6","schema_version":"1.0","canonical_sha256":"2bc87c697ee53cdeef426944ce0716a0b7c46eab3c9c23e19db666160686c4ba","source":{"kind":"arxiv","id":"2309.16349","version":2},"attestation_state":"computed","paper":{"title":"Human Feedback is not Gold Standard","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Max Bartolo, Phil Blunsom, Tom Hosking","submitted_at":"2023-09-28T11:18:20Z","abstract_excerpt":"Human feedback has become the de facto standard for evaluating the performance of Large Language Models, and is increasingly being used as a training objective. However, it is not clear which properties of a generated output this single `preference' score captures. We hypothesise that preference scores are subjective and open to undesirable biases. We critically analyse the use of human feedback for both training and evaluation, to verify whether it fully captures a range of crucial error criteria. We find that while preference scores have fairly good coverage, they under-represent important a"},"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":"2309.16349","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-28T11:18:20Z","cross_cats_sorted":[],"title_canon_sha256":"8011837a28a8116e44877fac6beefb7a2bce0352dbca11eddc02564e4e2829c8","abstract_canon_sha256":"24455077c2762b18b4bf23d73c69dd18ff0e43d992b1d1abbbbc54ca3a79f303"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:33:55.369929Z","signature_b64":"U9o6fbA4ccYYF1miPh91qPU/xvG59RVIu6kHpRJro6CzjX4VW+orKk4iHKbUxdsoO0Mp1C0AXD6bHcg0OC2WAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2bc87c697ee53cdeef426944ce0716a0b7c46eab3c9c23e19db666160686c4ba","last_reissued_at":"2026-07-05T07:33:55.369339Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:33:55.369339Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Human Feedback is not Gold Standard","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Max Bartolo, Phil Blunsom, Tom Hosking","submitted_at":"2023-09-28T11:18:20Z","abstract_excerpt":"Human feedback has become the de facto standard for evaluating the performance of Large Language Models, and is increasingly being used as a training objective. However, it is not clear which properties of a generated output this single `preference' score captures. We hypothesise that preference scores are subjective and open to undesirable biases. We critically analyse the use of human feedback for both training and evaluation, to verify whether it fully captures a range of crucial error criteria. We find that while preference scores have fairly good coverage, they under-represent important a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.16349","kind":"arxiv","version":2},"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/2309.16349/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":"2309.16349","created_at":"2026-07-05T07:33:55.369407+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.16349v2","created_at":"2026-07-05T07:33:55.369407+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.16349","created_at":"2026-07-05T07:33:55.369407+00:00"},{"alias_kind":"pith_short_12","alias_value":"FPEHY2L64U6N","created_at":"2026-07-05T07:33:55.369407+00:00"},{"alias_kind":"pith_short_16","alias_value":"FPEHY2L64U6N532C","created_at":"2026-07-05T07:33:55.369407+00:00"},{"alias_kind":"pith_short_8","alias_value":"FPEHY2L6","created_at":"2026-07-05T07:33:55.369407+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2402.05070","citing_title":"A Roadmap to Pluralistic Alignment","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC","json":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC.json","graph_json":"https://pith.science/api/pith-number/FPEHY2L64U6N532CNFCM4BYWUC/graph.json","events_json":"https://pith.science/api/pith-number/FPEHY2L64U6N532CNFCM4BYWUC/events.json","paper":"https://pith.science/paper/FPEHY2L6"},"agent_actions":{"view_html":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC","download_json":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC.json","view_paper":"https://pith.science/paper/FPEHY2L6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.16349&json=true","fetch_graph":"https://pith.science/api/pith-number/FPEHY2L64U6N532CNFCM4BYWUC/graph.json","fetch_events":"https://pith.science/api/pith-number/FPEHY2L64U6N532CNFCM4BYWUC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC/action/storage_attestation","attest_author":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC/action/author_attestation","sign_citation":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC/action/citation_signature","submit_replication":"https://pith.science/pith/FPEHY2L64U6N532CNFCM4BYWUC/action/replication_record"}},"created_at":"2026-07-05T07:33:55.369407+00:00","updated_at":"2026-07-05T07:33:55.369407+00:00"}