{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:SE6CXDTVQW7RCHIYPPJ4X7CHOY","short_pith_number":"pith:SE6CXDTV","schema_version":"1.0","canonical_sha256":"913c2b8e7585bf111d187bd3cbfc47763f8ea2d9853efa34cc9cff866a64d062","source":{"kind":"arxiv","id":"2607.08535","version":1},"attestation_state":"computed","paper":{"title":"When the Judge Changes, So Does the Measurement: Auditing LLM-as-Judge Reliability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Xiaokun Yang, Yinghan Hou, Zongyou Yang","submitted_at":"2026-07-09T14:31:05Z","abstract_excerpt":"An LLM-as-judge score can move even when the candidate responses stay fixed, simply because the evaluator has changed. We treat this evaluator-replacement ambiguity as a measurement-validity problem. Across four judgment datasets, we compare two upgrade paths available in practice: scaling Qwen3 dense judges from 1.7B to 32B parameters and moving across MiniMax M2-M2.7 released APIs. The main pattern is that judge upgrades are not interchangeable: only Qwen3 1.7B to 4B gives a robust adjacent gain, while MiniMax adjacent releases do not. Stronger judges reduce but do not remove position and ve"},"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":"2607.08535","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-07-09T14:31:05Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"125d40a3ea9128e6835674d02451e6e72457ae4ce0e86f59e86a679030ff552d","abstract_canon_sha256":"0b5e84caca08889b672c3677244a4751b45513eb3b3a31106a0456f4a4c307bb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-10T01:19:53.063681Z","signature_b64":"OKBSdTB+wHa9DYgNir5afbpKhRYUyyLi0vCLE5tWXv1PeiDeuizOFUo2kGCwtiEeacq6R0qpolAv+1rQmuFRCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"913c2b8e7585bf111d187bd3cbfc47763f8ea2d9853efa34cc9cff866a64d062","last_reissued_at":"2026-07-10T01:19:53.063266Z","signature_status":"signed_v1","first_computed_at":"2026-07-10T01:19:53.063266Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When the Judge Changes, So Does the Measurement: Auditing LLM-as-Judge Reliability","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Xiaokun Yang, Yinghan Hou, Zongyou Yang","submitted_at":"2026-07-09T14:31:05Z","abstract_excerpt":"An LLM-as-judge score can move even when the candidate responses stay fixed, simply because the evaluator has changed. We treat this evaluator-replacement ambiguity as a measurement-validity problem. Across four judgment datasets, we compare two upgrade paths available in practice: scaling Qwen3 dense judges from 1.7B to 32B parameters and moving across MiniMax M2-M2.7 released APIs. The main pattern is that judge upgrades are not interchangeable: only Qwen3 1.7B to 4B gives a robust adjacent gain, while MiniMax adjacent releases do not. Stronger judges reduce but do not remove position and ve"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.08535","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/2607.08535/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":"2607.08535","created_at":"2026-07-10T01:19:53.063323+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.08535v1","created_at":"2026-07-10T01:19:53.063323+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.08535","created_at":"2026-07-10T01:19:53.063323+00:00"},{"alias_kind":"pith_short_12","alias_value":"SE6CXDTVQW7R","created_at":"2026-07-10T01:19:53.063323+00:00"},{"alias_kind":"pith_short_16","alias_value":"SE6CXDTVQW7RCHIY","created_at":"2026-07-10T01:19:53.063323+00:00"},{"alias_kind":"pith_short_8","alias_value":"SE6CXDTV","created_at":"2026-07-10T01:19:53.063323+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.24268","citing_title":"Accuracy Hides How Language Models Fail: Measuring Failure States Under Matched Output Budgets","ref_index":7,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY","json":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY.json","graph_json":"https://pith.science/api/pith-number/SE6CXDTVQW7RCHIYPPJ4X7CHOY/graph.json","events_json":"https://pith.science/api/pith-number/SE6CXDTVQW7RCHIYPPJ4X7CHOY/events.json","paper":"https://pith.science/paper/SE6CXDTV"},"agent_actions":{"view_html":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY","download_json":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY.json","view_paper":"https://pith.science/paper/SE6CXDTV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.08535&json=true","fetch_graph":"https://pith.science/api/pith-number/SE6CXDTVQW7RCHIYPPJ4X7CHOY/graph.json","fetch_events":"https://pith.science/api/pith-number/SE6CXDTVQW7RCHIYPPJ4X7CHOY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY/action/storage_attestation","attest_author":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY/action/author_attestation","sign_citation":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY/action/citation_signature","submit_replication":"https://pith.science/pith/SE6CXDTVQW7RCHIYPPJ4X7CHOY/action/replication_record"}},"created_at":"2026-07-10T01:19:53.063323+00:00","updated_at":"2026-07-10T01:19:53.063323+00:00"}