{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CYEJOF4COFNCLH63JLJN5RUYQE","short_pith_number":"pith:CYEJOF4C","schema_version":"1.0","canonical_sha256":"1608971782715a259fdb4ad2dec6988135fe6fb718cee313ad3ba3342e04346a","source":{"kind":"arxiv","id":"2404.12318","version":2},"attestation_state":"computed","paper":{"title":"Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ahmad Beirami, Ananth Balashankar, Jacob Eisenstein, Yoon Kim, Zhaofeng Wu","submitted_at":"2024-04-18T16:52:36Z","abstract_excerpt":"Aligning language models (LMs) based on human-annotated preference data is a crucial step in obtaining practical and performant LM-based systems. However, multilingual human preference data are difficult to obtain at scale, making it challenging to extend this framework to diverse languages. In this work, we evaluate a simple approach for zero-shot cross-lingual alignment, where a reward model is trained on preference data in one source language and directly applied to other target languages. On summarization and open-ended dialog generation, we show that this method is consistently successful"},"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":"2404.12318","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-04-18T16:52:36Z","cross_cats_sorted":[],"title_canon_sha256":"60e34aa50e8a763f96fff37105688a5d820e2937cd3175c04e9f9bbbda009b22","abstract_canon_sha256":"3fc7025037d23ebf01772cda13cf92b6af637830a9cec6602a17cf722d7cfdd6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:19:55.346007Z","signature_b64":"ET6tqy2C6qhEsnJbJ73+WxI25na6nhK4xuqVX4wNOPRwH6c14fg/iKYiNVdyWkVjmE5FMXtnjo2MNiDIuzTmBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1608971782715a259fdb4ad2dec6988135fe6fb718cee313ad3ba3342e04346a","last_reissued_at":"2026-07-05T09:19:55.345561Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:19:55.345561Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reuse Your Rewards: Reward Model Transfer for Zero-Shot Cross-Lingual Alignment","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ahmad Beirami, Ananth Balashankar, Jacob Eisenstein, Yoon Kim, Zhaofeng Wu","submitted_at":"2024-04-18T16:52:36Z","abstract_excerpt":"Aligning language models (LMs) based on human-annotated preference data is a crucial step in obtaining practical and performant LM-based systems. However, multilingual human preference data are difficult to obtain at scale, making it challenging to extend this framework to diverse languages. In this work, we evaluate a simple approach for zero-shot cross-lingual alignment, where a reward model is trained on preference data in one source language and directly applied to other target languages. On summarization and open-ended dialog generation, we show that this method is consistently successful"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.12318","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/2404.12318/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":"2404.12318","created_at":"2026-07-05T09:19:55.345619+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.12318v2","created_at":"2026-07-05T09:19:55.345619+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.12318","created_at":"2026-07-05T09:19:55.345619+00:00"},{"alias_kind":"pith_short_12","alias_value":"CYEJOF4COFNC","created_at":"2026-07-05T09:19:55.345619+00:00"},{"alias_kind":"pith_short_16","alias_value":"CYEJOF4COFNCLH63","created_at":"2026-07-05T09:19:55.345619+00:00"},{"alias_kind":"pith_short_8","alias_value":"CYEJOF4C","created_at":"2026-07-05T09:19:55.345619+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.00911","citing_title":"Aligning LLMs with Domain Invariant Reward Models","ref_index":46,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE","json":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE.json","graph_json":"https://pith.science/api/pith-number/CYEJOF4COFNCLH63JLJN5RUYQE/graph.json","events_json":"https://pith.science/api/pith-number/CYEJOF4COFNCLH63JLJN5RUYQE/events.json","paper":"https://pith.science/paper/CYEJOF4C"},"agent_actions":{"view_html":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE","download_json":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE.json","view_paper":"https://pith.science/paper/CYEJOF4C","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.12318&json=true","fetch_graph":"https://pith.science/api/pith-number/CYEJOF4COFNCLH63JLJN5RUYQE/graph.json","fetch_events":"https://pith.science/api/pith-number/CYEJOF4COFNCLH63JLJN5RUYQE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE/action/storage_attestation","attest_author":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE/action/author_attestation","sign_citation":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE/action/citation_signature","submit_replication":"https://pith.science/pith/CYEJOF4COFNCLH63JLJN5RUYQE/action/replication_record"}},"created_at":"2026-07-05T09:19:55.345619+00:00","updated_at":"2026-07-05T09:19:55.345619+00:00"}