{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:D444J7UARXO4NGKORPDJXOSPFP","short_pith_number":"pith:D444J7UA","schema_version":"1.0","canonical_sha256":"1f39c4fe808dddc6994e8bc69bba4f2bf87d9dd2ba366e11645d8d9b1b948629","source":{"kind":"arxiv","id":"2503.19877","version":2},"attestation_state":"computed","paper":{"title":"Scaling Evaluation-time Compute with Reasoning Models as Evaluators","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Carolin Lawrence, Graham Neubig, Ian Wu, Jinu Lee, Julia Hockenmaier, Kiril Gashteovski, Mingyeong Moon, Sean Welleck, Seongyun Lee, Seungone Kim, Xiang Yue","submitted_at":"2025-03-25T17:41:18Z","abstract_excerpt":"As language model (LM) outputs get more and more natural, it is becoming more difficult than ever to evaluate their quality. Simultaneously, increasing LMs' \"thinking\" time through scaling test-time compute has proven an effective technique to solve challenging problems in domains such as math and code. This raises a natural question: can an LM's evaluation capability also be improved by spending more test-time compute? To answer this, we investigate employing reasoning models-LMs that natively generate long chain-of-thought reasoning-as evaluators. Specifically, we examine methods to leverage"},"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":"2503.19877","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-25T17:41:18Z","cross_cats_sorted":[],"title_canon_sha256":"517380c753e75f8492a57a55e7719328ee353b3743d1fc637bbcf6f7c8518e51","abstract_canon_sha256":"2d2d17586029d92508780e41d6179af1276293e1c9eebfde567ada5791cef27d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:04:53.265219Z","signature_b64":"XAydjeMwk1+URPSlY4fV4tvdWiTM54NS6VAYWuZlcme8GPfgXoI6p385Fn8xqyneQhvLXkWRtNbtPrpYrMD3DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1f39c4fe808dddc6994e8bc69bba4f2bf87d9dd2ba366e11645d8d9b1b948629","last_reissued_at":"2026-05-20T01:04:53.264440Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:04:53.264440Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Evaluation-time Compute with Reasoning Models as Evaluators","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Carolin Lawrence, Graham Neubig, Ian Wu, Jinu Lee, Julia Hockenmaier, Kiril Gashteovski, Mingyeong Moon, Sean Welleck, Seongyun Lee, Seungone Kim, Xiang Yue","submitted_at":"2025-03-25T17:41:18Z","abstract_excerpt":"As language model (LM) outputs get more and more natural, it is becoming more difficult than ever to evaluate their quality. Simultaneously, increasing LMs' \"thinking\" time through scaling test-time compute has proven an effective technique to solve challenging problems in domains such as math and code. This raises a natural question: can an LM's evaluation capability also be improved by spending more test-time compute? To answer this, we investigate employing reasoning models-LMs that natively generate long chain-of-thought reasoning-as evaluators. Specifically, we examine methods to leverage"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.19877","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/2503.19877/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":"2503.19877","created_at":"2026-05-20T01:04:53.264538+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.19877v2","created_at":"2026-05-20T01:04:53.264538+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.19877","created_at":"2026-05-20T01:04:53.264538+00:00"},{"alias_kind":"pith_short_12","alias_value":"D444J7UARXO4","created_at":"2026-05-20T01:04:53.264538+00:00"},{"alias_kind":"pith_short_16","alias_value":"D444J7UARXO4NGKO","created_at":"2026-05-20T01:04:53.264538+00:00"},{"alias_kind":"pith_short_8","alias_value":"D444J7UA","created_at":"2026-05-20T01:04:53.264538+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2605.15529","citing_title":"Process Rewards with Learned Reliability","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2509.09438","citing_title":"GrACE: A Generative Approach to Better Confidence Elicitation and Efficient Test-Time Scaling in Large Language Models","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2509.11206","citing_title":"Evalet: Evaluating Large Language Models through Functional Fragmentation","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2509.23542","citing_title":"On the Shelf Life of Fine-Tuned LLM-Judges: Future-Proofing, Backward-Compatibility, and Question Generalization","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP","json":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP.json","graph_json":"https://pith.science/api/pith-number/D444J7UARXO4NGKORPDJXOSPFP/graph.json","events_json":"https://pith.science/api/pith-number/D444J7UARXO4NGKORPDJXOSPFP/events.json","paper":"https://pith.science/paper/D444J7UA"},"agent_actions":{"view_html":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP","download_json":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP.json","view_paper":"https://pith.science/paper/D444J7UA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.19877&json=true","fetch_graph":"https://pith.science/api/pith-number/D444J7UARXO4NGKORPDJXOSPFP/graph.json","fetch_events":"https://pith.science/api/pith-number/D444J7UARXO4NGKORPDJXOSPFP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP/action/storage_attestation","attest_author":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP/action/author_attestation","sign_citation":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP/action/citation_signature","submit_replication":"https://pith.science/pith/D444J7UARXO4NGKORPDJXOSPFP/action/replication_record"}},"created_at":"2026-05-20T01:04:53.264538+00:00","updated_at":"2026-05-20T01:04:53.264538+00:00"}