{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KERZ3COJQGIOIJFCAS6QLVAXP3","short_pith_number":"pith:KERZ3COJ","schema_version":"1.0","canonical_sha256":"51239d89c98190e424a204bd05d4177ec8ce2bed1bb07196604fe9364e01a29e","source":{"kind":"arxiv","id":"2607.14480","version":1},"attestation_state":"computed","paper":{"title":"LLM Evaluators are Biased across Languages","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Anna Korhonen, Ej Zhou, Lucas Resck, Zheng Hui","submitted_at":"2026-07-16T01:53:09Z","abstract_excerpt":"LLM evaluators (trained reward models and prompted LLM-as-a-Judge) are routinely validated via pairwise accuracy. In a multilingual setting, this operates under the premise that high pairwise accuracy implies reliable, language-neutral scoring. We show that this assumption does not hold. We conduct experiments with semantically identical instruction-response pairs across 23 languages, and find that multilingual evaluators assign significantly different scores to different evaluation languages. The bias is statistically significant and consistent across eight open-weight evaluators of different"},"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":"2607.14480","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-07-16T01:53:09Z","cross_cats_sorted":[],"title_canon_sha256":"acd776553876caa24ebec736039a1724e01bac111b422f293f60eb87deb4c56f","abstract_canon_sha256":"3e74ef9c59484a32c4a8144ead1a37be3602d1ba9976cfe33cc6bb1385ec4756"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-17T00:21:14.627116Z","signature_b64":"ET4zibiUQolQLB7831awVrJ/0vpgyJGpbhLkhr+v6XnYNerYtYrPakSW8cSc+FgQ6tjtSBjtx1iaiZIAJplvCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"51239d89c98190e424a204bd05d4177ec8ce2bed1bb07196604fe9364e01a29e","last_reissued_at":"2026-07-17T00:21:14.626255Z","signature_status":"signed_v1","first_computed_at":"2026-07-17T00:21:14.626255Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLM Evaluators are Biased across Languages","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Anna Korhonen, Ej Zhou, Lucas Resck, Zheng Hui","submitted_at":"2026-07-16T01:53:09Z","abstract_excerpt":"LLM evaluators (trained reward models and prompted LLM-as-a-Judge) are routinely validated via pairwise accuracy. In a multilingual setting, this operates under the premise that high pairwise accuracy implies reliable, language-neutral scoring. We show that this assumption does not hold. We conduct experiments with semantically identical instruction-response pairs across 23 languages, and find that multilingual evaluators assign significantly different scores to different evaluation languages. The bias is statistically significant and consistent across eight open-weight evaluators of different"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.14480","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/2607.14480/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":"2607.14480","created_at":"2026-07-17T00:21:14.626712+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.14480v1","created_at":"2026-07-17T00:21:14.626712+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.14480","created_at":"2026-07-17T00:21:14.626712+00:00"},{"alias_kind":"pith_short_12","alias_value":"KERZ3COJQGIO","created_at":"2026-07-17T00:21:14.626712+00:00"},{"alias_kind":"pith_short_16","alias_value":"KERZ3COJQGIOIJFC","created_at":"2026-07-17T00:21:14.626712+00:00"},{"alias_kind":"pith_short_8","alias_value":"KERZ3COJ","created_at":"2026-07-17T00:21:14.626712+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/KERZ3COJQGIOIJFCAS6QLVAXP3","json":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3.json","graph_json":"https://pith.science/api/pith-number/KERZ3COJQGIOIJFCAS6QLVAXP3/graph.json","events_json":"https://pith.science/api/pith-number/KERZ3COJQGIOIJFCAS6QLVAXP3/events.json","paper":"https://pith.science/paper/KERZ3COJ"},"agent_actions":{"view_html":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3","download_json":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3.json","view_paper":"https://pith.science/paper/KERZ3COJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.14480&json=true","fetch_graph":"https://pith.science/api/pith-number/KERZ3COJQGIOIJFCAS6QLVAXP3/graph.json","fetch_events":"https://pith.science/api/pith-number/KERZ3COJQGIOIJFCAS6QLVAXP3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3/action/storage_attestation","attest_author":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3/action/author_attestation","sign_citation":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3/action/citation_signature","submit_replication":"https://pith.science/pith/KERZ3COJQGIOIJFCAS6QLVAXP3/action/replication_record"}},"created_at":"2026-07-17T00:21:14.626712+00:00","updated_at":"2026-07-17T00:21:14.626712+00:00"}