{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FNHXEPP2X66YFRO3BYI2ORCCAL","short_pith_number":"pith:FNHXEPP2","schema_version":"1.0","canonical_sha256":"2b4f723dfabfbd82c5db0e11a7444202e5577670c1e5a10863d0e0b697f3eb16","source":{"kind":"arxiv","id":"2505.12301","version":1},"attestation_state":"computed","paper":{"title":"Beyond Single-Point Judgment: Distribution Alignment for LLM-as-a-Judge","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Haoran Tan, Hao Yang, Luyu Chen, Quanyu Dai, Xu Chen, Zeyu Zhang, Zhenhua Dong","submitted_at":"2025-05-18T08:33:09Z","abstract_excerpt":"LLMs have emerged as powerful evaluators in the LLM-as-a-Judge paradigm, offering significant efficiency and flexibility compared to human judgments. However, previous methods primarily rely on single-point evaluations, overlooking the inherent diversity and uncertainty in human evaluations. This approach leads to information loss and decreases the reliability of evaluations. To address this limitation, we propose a novel training framework that explicitly aligns the LLM-generated judgment distribution with empirical human distributions. Specifically, we propose a distributional alignment obje"},"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":"2505.12301","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-18T08:33:09Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"37f4340872aef625e3a9b6258897640566c0d946503fcf333534d5318cd92d91","abstract_canon_sha256":"8c047b75f4e575d3d1bd4f4e6dc36ce16b43c1be27ee46b26f3595a2a8207650"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:01.831142Z","signature_b64":"XM+ac5RFy8r/1WQfTlmGRucHoGGdfwku8ZdHz+RyR9fQFMQUeWlCFt1RU5faDmTjgVelrtBay/mfn9gyGsMzAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2b4f723dfabfbd82c5db0e11a7444202e5577670c1e5a10863d0e0b697f3eb16","last_reissued_at":"2026-07-05T11:05:01.830638Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:01.830638Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Single-Point Judgment: Distribution Alignment for LLM-as-a-Judge","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.AI","authors_text":"Haoran Tan, Hao Yang, Luyu Chen, Quanyu Dai, Xu Chen, Zeyu Zhang, Zhenhua Dong","submitted_at":"2025-05-18T08:33:09Z","abstract_excerpt":"LLMs have emerged as powerful evaluators in the LLM-as-a-Judge paradigm, offering significant efficiency and flexibility compared to human judgments. However, previous methods primarily rely on single-point evaluations, overlooking the inherent diversity and uncertainty in human evaluations. This approach leads to information loss and decreases the reliability of evaluations. To address this limitation, we propose a novel training framework that explicitly aligns the LLM-generated judgment distribution with empirical human distributions. Specifically, we propose a distributional alignment obje"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.12301","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/2505.12301/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":"2505.12301","created_at":"2026-07-05T11:05:01.830701+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.12301v1","created_at":"2026-07-05T11:05:01.830701+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.12301","created_at":"2026-07-05T11:05:01.830701+00:00"},{"alias_kind":"pith_short_12","alias_value":"FNHXEPP2X66Y","created_at":"2026-07-05T11:05:01.830701+00:00"},{"alias_kind":"pith_short_16","alias_value":"FNHXEPP2X66YFRO3","created_at":"2026-07-05T11:05:01.830701+00:00"},{"alias_kind":"pith_short_8","alias_value":"FNHXEPP2","created_at":"2026-07-05T11:05:01.830701+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.10908","citing_title":"Order Matters: LVLMs as Judges for Temporal Reasoning in Image Sequences","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL","json":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL.json","graph_json":"https://pith.science/api/pith-number/FNHXEPP2X66YFRO3BYI2ORCCAL/graph.json","events_json":"https://pith.science/api/pith-number/FNHXEPP2X66YFRO3BYI2ORCCAL/events.json","paper":"https://pith.science/paper/FNHXEPP2"},"agent_actions":{"view_html":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL","download_json":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL.json","view_paper":"https://pith.science/paper/FNHXEPP2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.12301&json=true","fetch_graph":"https://pith.science/api/pith-number/FNHXEPP2X66YFRO3BYI2ORCCAL/graph.json","fetch_events":"https://pith.science/api/pith-number/FNHXEPP2X66YFRO3BYI2ORCCAL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL/action/storage_attestation","attest_author":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL/action/author_attestation","sign_citation":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL/action/citation_signature","submit_replication":"https://pith.science/pith/FNHXEPP2X66YFRO3BYI2ORCCAL/action/replication_record"}},"created_at":"2026-07-05T11:05:01.830701+00:00","updated_at":"2026-07-05T11:05:01.830701+00:00"}