{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4XORNMDX4XDLNJ4XM7NCJLVTLP","short_pith_number":"pith:4XORNMDX","schema_version":"1.0","canonical_sha256":"e5dd16b077e5c6b6a79767da24aeb35bf6d622833f9ba1e5db94eb2c04ae1738","source":{"kind":"arxiv","id":"2410.19133","version":5},"attestation_state":"computed","paper":{"title":"Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Faeze Brahman, Hannaneh Hajishirzi, Lester James V. Miranda, Noah A. Smith, Pradeep Dasigi, Sachin Kumar, Valentina Pyatkin, Yanai Elazar, Yizhong Wang","submitted_at":"2024-10-24T20:04:15Z","abstract_excerpt":"Learning from human feedback has enabled the alignment of language models (LMs) with human preferences. However, collecting human preferences is expensive and time-consuming, with highly variable annotation quality. An appealing alternative is to distill preferences from LMs as a source of synthetic annotations, offering a cost-effective and scalable alternative, albeit susceptible to other biases and errors. In this work, we introduce HyPER, a Hybrid Preference routER that defers an annotation to either humans or LMs, achieving better annotation quality while reducing the cost of human-only 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":"2410.19133","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-10-24T20:04:15Z","cross_cats_sorted":[],"title_canon_sha256":"1f4a856f790974a0488aeb9961c8b6b2e322ca198c3ffa4f6ad3cc616e1408b7","abstract_canon_sha256":"1a786a74839055d053fe6346b340cfdd966d4d771cba8da6ea0c58464381708e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:14.814005Z","signature_b64":"R/MLXxifESuugdCGRyEkhJ0QqRwNMgoWiUS7l1KHxcJ7Dyn4Jr0amT2juSGWchYovyQJrsjdwQUMMeh4QbbmAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e5dd16b077e5c6b6a79767da24aeb35bf6d622833f9ba1e5db94eb2c04ae1738","last_reissued_at":"2026-07-05T11:13:14.813494Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:14.813494Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hybrid Preferences: Learning to Route Instances for Human vs. AI Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Faeze Brahman, Hannaneh Hajishirzi, Lester James V. Miranda, Noah A. Smith, Pradeep Dasigi, Sachin Kumar, Valentina Pyatkin, Yanai Elazar, Yizhong Wang","submitted_at":"2024-10-24T20:04:15Z","abstract_excerpt":"Learning from human feedback has enabled the alignment of language models (LMs) with human preferences. However, collecting human preferences is expensive and time-consuming, with highly variable annotation quality. An appealing alternative is to distill preferences from LMs as a source of synthetic annotations, offering a cost-effective and scalable alternative, albeit susceptible to other biases and errors. In this work, we introduce HyPER, a Hybrid Preference routER that defers an annotation to either humans or LMs, achieving better annotation quality while reducing the cost of human-only a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.19133","kind":"arxiv","version":5},"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/2410.19133/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":"2410.19133","created_at":"2026-07-05T11:13:14.813557+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.19133v5","created_at":"2026-07-05T11:13:14.813557+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.19133","created_at":"2026-07-05T11:13:14.813557+00:00"},{"alias_kind":"pith_short_12","alias_value":"4XORNMDX4XDL","created_at":"2026-07-05T11:13:14.813557+00:00"},{"alias_kind":"pith_short_16","alias_value":"4XORNMDX4XDLNJ4X","created_at":"2026-07-05T11:13:14.813557+00:00"},{"alias_kind":"pith_short_8","alias_value":"4XORNMDX","created_at":"2026-07-05T11:13:14.813557+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27578","citing_title":"PEBS: Per-rater Empirical-Bayes Shrinkage for RLHF Reward-Model Calibration","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP","json":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP.json","graph_json":"https://pith.science/api/pith-number/4XORNMDX4XDLNJ4XM7NCJLVTLP/graph.json","events_json":"https://pith.science/api/pith-number/4XORNMDX4XDLNJ4XM7NCJLVTLP/events.json","paper":"https://pith.science/paper/4XORNMDX"},"agent_actions":{"view_html":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP","download_json":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP.json","view_paper":"https://pith.science/paper/4XORNMDX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.19133&json=true","fetch_graph":"https://pith.science/api/pith-number/4XORNMDX4XDLNJ4XM7NCJLVTLP/graph.json","fetch_events":"https://pith.science/api/pith-number/4XORNMDX4XDLNJ4XM7NCJLVTLP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP/action/storage_attestation","attest_author":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP/action/author_attestation","sign_citation":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP/action/citation_signature","submit_replication":"https://pith.science/pith/4XORNMDX4XDLNJ4XM7NCJLVTLP/action/replication_record"}},"created_at":"2026-07-05T11:13:14.813557+00:00","updated_at":"2026-07-05T11:13:14.813557+00:00"}