{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:O7BVHENPAKGK5J4P24JJPQUP5Z","short_pith_number":"pith:O7BVHENP","schema_version":"1.0","canonical_sha256":"77c35391af028caea78fd71297c28fee73c52904afb9926e8373c3d042479ae2","source":{"kind":"arxiv","id":"2502.10563","version":2},"attestation_state":"computed","paper":{"title":"Accelerating Unbiased LLM Evaluation via Synthetic Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Andrea Zanette, Yuda Song, Zhaoyi Zhou","submitted_at":"2025-02-14T21:27:09Z","abstract_excerpt":"When developing new large language models (LLMs), a key step is evaluating their final performance, often by computing the win-rate against a reference model based on external feedback. Human feedback is the gold standard, particularly for capturing nuanced qualities like coherence, readability, and alignment with human expectations. However, human evaluations are costly -- even for large tech companies -- and when conducted with active users, they may negatively impact user experience. A promising alternative is synthetic feedback, where evaluations are conducted by other large language model"},"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":"2502.10563","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-14T21:27:09Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"436d421ceec9c3beb5b85ca58ef8701e4aa2c410067c46bdb5be1601ea68ff0d","abstract_canon_sha256":"d4440f477c070256a65c32c14c2bc4ce91a2704445bd38a5ec4196325634592d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:19:37.905615Z","signature_b64":"EPe7FHnNS81O9JspcGnUeypA4A6juh3J2Lk9a0GFSQbHZGPY1fMWrOOoHeKCuOLxbUE9QEMPwBUMP1zVAvKpBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"77c35391af028caea78fd71297c28fee73c52904afb9926e8373c3d042479ae2","last_reissued_at":"2026-07-05T10:19:37.905123Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:19:37.905123Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accelerating Unbiased LLM Evaluation via Synthetic Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Andrea Zanette, Yuda Song, Zhaoyi Zhou","submitted_at":"2025-02-14T21:27:09Z","abstract_excerpt":"When developing new large language models (LLMs), a key step is evaluating their final performance, often by computing the win-rate against a reference model based on external feedback. Human feedback is the gold standard, particularly for capturing nuanced qualities like coherence, readability, and alignment with human expectations. However, human evaluations are costly -- even for large tech companies -- and when conducted with active users, they may negatively impact user experience. A promising alternative is synthetic feedback, where evaluations are conducted by other large language model"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.10563","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/2502.10563/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":"2502.10563","created_at":"2026-07-05T10:19:37.905181+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.10563v2","created_at":"2026-07-05T10:19:37.905181+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.10563","created_at":"2026-07-05T10:19:37.905181+00:00"},{"alias_kind":"pith_short_12","alias_value":"O7BVHENPAKGK","created_at":"2026-07-05T10:19:37.905181+00:00"},{"alias_kind":"pith_short_16","alias_value":"O7BVHENPAKGK5J4P","created_at":"2026-07-05T10:19:37.905181+00:00"},{"alias_kind":"pith_short_8","alias_value":"O7BVHENP","created_at":"2026-07-05T10:19:37.905181+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.05159","citing_title":"X4Val: Learning Neural Surrogates for Variance-Reduced Policy Evaluation","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10075","citing_title":"Active Testing of Large Language Models via Approximate Neyman Allocation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10075","citing_title":"Active Testing of Large Language Models via Approximate Neyman Allocation","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z","json":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z.json","graph_json":"https://pith.science/api/pith-number/O7BVHENPAKGK5J4P24JJPQUP5Z/graph.json","events_json":"https://pith.science/api/pith-number/O7BVHENPAKGK5J4P24JJPQUP5Z/events.json","paper":"https://pith.science/paper/O7BVHENP"},"agent_actions":{"view_html":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z","download_json":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z.json","view_paper":"https://pith.science/paper/O7BVHENP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.10563&json=true","fetch_graph":"https://pith.science/api/pith-number/O7BVHENPAKGK5J4P24JJPQUP5Z/graph.json","fetch_events":"https://pith.science/api/pith-number/O7BVHENPAKGK5J4P24JJPQUP5Z/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z/action/timestamp_anchor","attest_storage":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z/action/storage_attestation","attest_author":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z/action/author_attestation","sign_citation":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z/action/citation_signature","submit_replication":"https://pith.science/pith/O7BVHENPAKGK5J4P24JJPQUP5Z/action/replication_record"}},"created_at":"2026-07-05T10:19:37.905181+00:00","updated_at":"2026-07-05T10:19:37.905181+00:00"}