{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KDXE7KW6JUFO6DJDBLSV4SBMYO","short_pith_number":"pith:KDXE7KW6","schema_version":"1.0","canonical_sha256":"50ee4faade4d0aef0d230ae55e482cc3949902acebaed906184a39a571180631","source":{"kind":"arxiv","id":"2606.19607","version":1},"attestation_state":"computed","paper":{"title":"Which Pairs to Compare for LLM Post-Training?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.AP"],"primary_cat":"cs.AI","authors_text":"Jiangze Han, Vineet Goyal, Will Ma","submitted_at":"2026-06-17T21:19:01Z","abstract_excerpt":"Preference-based post-training has become a central paradigm for aligning language models. A common data-collection strategy is to generate a small set of completions for each prompt and label the resulting comparison pairs. However, human preference labels are often much more expensive than generating additional completions, suggesting a different use of the same labeling budget: generate a larger pool of completions, but label only the most informative comparison pairs. This paper studies which pairs should be compared in preference-based post-training. We formulate comparison curation as 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":"2606.19607","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-06-17T21:19:01Z","cross_cats_sorted":["stat.AP"],"title_canon_sha256":"58eef06d36970d9fd81cdcfdf99903d2623a991f586fc7ff8aa34493655d0ea4","abstract_canon_sha256":"04516e0854eeee81092542259dc18740c127a907fa55e910bf3c6681600088a9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:12:30.053176Z","signature_b64":"hryOw/uBDQT3QUlcc9ygGXFgZdz1GA87qEV4IPqsd9i7gw0m2Iyfizca+62fuwr9tm1sGK54ZEJHjn1jMQk3Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"50ee4faade4d0aef0d230ae55e482cc3949902acebaed906184a39a571180631","last_reissued_at":"2026-06-19T16:12:30.052837Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:12:30.052837Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Which Pairs to Compare for LLM Post-Training?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.AP"],"primary_cat":"cs.AI","authors_text":"Jiangze Han, Vineet Goyal, Will Ma","submitted_at":"2026-06-17T21:19:01Z","abstract_excerpt":"Preference-based post-training has become a central paradigm for aligning language models. A common data-collection strategy is to generate a small set of completions for each prompt and label the resulting comparison pairs. However, human preference labels are often much more expensive than generating additional completions, suggesting a different use of the same labeling budget: generate a larger pool of completions, but label only the most informative comparison pairs. This paper studies which pairs should be compared in preference-based post-training. We formulate comparison curation as a "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.19607","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/2606.19607/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":"2606.19607","created_at":"2026-06-19T16:12:30.052898+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.19607v1","created_at":"2026-06-19T16:12:30.052898+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.19607","created_at":"2026-06-19T16:12:30.052898+00:00"},{"alias_kind":"pith_short_12","alias_value":"KDXE7KW6JUFO","created_at":"2026-06-19T16:12:30.052898+00:00"},{"alias_kind":"pith_short_16","alias_value":"KDXE7KW6JUFO6DJD","created_at":"2026-06-19T16:12:30.052898+00:00"},{"alias_kind":"pith_short_8","alias_value":"KDXE7KW6","created_at":"2026-06-19T16:12:30.052898+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/KDXE7KW6JUFO6DJDBLSV4SBMYO","json":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO.json","graph_json":"https://pith.science/api/pith-number/KDXE7KW6JUFO6DJDBLSV4SBMYO/graph.json","events_json":"https://pith.science/api/pith-number/KDXE7KW6JUFO6DJDBLSV4SBMYO/events.json","paper":"https://pith.science/paper/KDXE7KW6"},"agent_actions":{"view_html":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO","download_json":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO.json","view_paper":"https://pith.science/paper/KDXE7KW6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.19607&json=true","fetch_graph":"https://pith.science/api/pith-number/KDXE7KW6JUFO6DJDBLSV4SBMYO/graph.json","fetch_events":"https://pith.science/api/pith-number/KDXE7KW6JUFO6DJDBLSV4SBMYO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO/action/storage_attestation","attest_author":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO/action/author_attestation","sign_citation":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO/action/citation_signature","submit_replication":"https://pith.science/pith/KDXE7KW6JUFO6DJDBLSV4SBMYO/action/replication_record"}},"created_at":"2026-06-19T16:12:30.052898+00:00","updated_at":"2026-06-19T16:12:30.052898+00:00"}