{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5TI55IC5KYTISG7PS2KJISJ56D","short_pith_number":"pith:5TI55IC5","schema_version":"1.0","canonical_sha256":"ecd1dea05d5626891bef969494493df0c13b9c2d1e66decd8fc9b641b4d9700a","source":{"kind":"arxiv","id":"2606.06586","version":1},"attestation_state":"computed","paper":{"title":"Improving Cross-Lingual Factual Recall via Consistency-Driven Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Eduardo Sanchez, Ektor Oikonomidis Doumpas, Eleftheria Kolokytha, George Burgess, Harry O'Donnell, Jonathan von Rad, Louis Arts, Pontus Stenetorp, Yao Lu","submitted_at":"2026-06-04T18:00:02Z","abstract_excerpt":"Large language models (LLMs) trained predominantly on English data encode substantial world knowledge, yet often fail to express it reliably in other languages, a phenomenon known as cross-lingual factual inconsistency. To study and address this, we introduce PolyFact, a large-scale parallel multilingual factual QA dataset containing 100K Wikidata-grounded facts across 12 typologically diverse languages. Using PolyFact, we compare light continual pretraining (CPT), supervised fine-tuning (SFT), and reinforcement learning via Group Relative Policy Optimization (GRPO) for improving cross-lingual"},"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":"2606.06586","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-04T18:00:02Z","cross_cats_sorted":[],"title_canon_sha256":"82ee735bda8416d9e176a60ccc7f84d8f998b0f90a65ad8b1dbbee0bd28f4ae9","abstract_canon_sha256":"6eaf802ac5b8e0bfd3269686b0addeb50b6637a4d64fcc3db540fb83a6f36d41"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T00:03:46.048605Z","signature_b64":"slQUpkA33rsikO1tk4Zt2uzOHp79nJsPqPWBnGqYyqNfegoPAASHPKSsSQUpspR/xY4qrh6Ilrw+j3B0yp7DCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ecd1dea05d5626891bef969494493df0c13b9c2d1e66decd8fc9b641b4d9700a","last_reissued_at":"2026-06-08T00:03:46.048059Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T00:03:46.048059Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving Cross-Lingual Factual Recall via Consistency-Driven Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Eduardo Sanchez, Ektor Oikonomidis Doumpas, Eleftheria Kolokytha, George Burgess, Harry O'Donnell, Jonathan von Rad, Louis Arts, Pontus Stenetorp, Yao Lu","submitted_at":"2026-06-04T18:00:02Z","abstract_excerpt":"Large language models (LLMs) trained predominantly on English data encode substantial world knowledge, yet often fail to express it reliably in other languages, a phenomenon known as cross-lingual factual inconsistency. To study and address this, we introduce PolyFact, a large-scale parallel multilingual factual QA dataset containing 100K Wikidata-grounded facts across 12 typologically diverse languages. Using PolyFact, we compare light continual pretraining (CPT), supervised fine-tuning (SFT), and reinforcement learning via Group Relative Policy Optimization (GRPO) for improving cross-lingual"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.06586","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/2606.06586/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":"2606.06586","created_at":"2026-06-08T00:03:46.048143+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.06586v1","created_at":"2026-06-08T00:03:46.048143+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.06586","created_at":"2026-06-08T00:03:46.048143+00:00"},{"alias_kind":"pith_short_12","alias_value":"5TI55IC5KYTI","created_at":"2026-06-08T00:03:46.048143+00:00"},{"alias_kind":"pith_short_16","alias_value":"5TI55IC5KYTISG7P","created_at":"2026-06-08T00:03:46.048143+00:00"},{"alias_kind":"pith_short_8","alias_value":"5TI55IC5","created_at":"2026-06-08T00:03:46.048143+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/5TI55IC5KYTISG7PS2KJISJ56D","json":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D.json","graph_json":"https://pith.science/api/pith-number/5TI55IC5KYTISG7PS2KJISJ56D/graph.json","events_json":"https://pith.science/api/pith-number/5TI55IC5KYTISG7PS2KJISJ56D/events.json","paper":"https://pith.science/paper/5TI55IC5"},"agent_actions":{"view_html":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D","download_json":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D.json","view_paper":"https://pith.science/paper/5TI55IC5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.06586&json=true","fetch_graph":"https://pith.science/api/pith-number/5TI55IC5KYTISG7PS2KJISJ56D/graph.json","fetch_events":"https://pith.science/api/pith-number/5TI55IC5KYTISG7PS2KJISJ56D/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D/action/storage_attestation","attest_author":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D/action/author_attestation","sign_citation":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D/action/citation_signature","submit_replication":"https://pith.science/pith/5TI55IC5KYTISG7PS2KJISJ56D/action/replication_record"}},"created_at":"2026-06-08T00:03:46.048143+00:00","updated_at":"2026-06-08T00:03:46.048143+00:00"}