{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RCQJTJEOT3OAC2UV2YLR2JJPU7","short_pith_number":"pith:RCQJTJEO","schema_version":"1.0","canonical_sha256":"88a099a48e9edc016a95d6171d252fa7e4b598faeb2847d8bc881e561c9b1538","source":{"kind":"arxiv","id":"2605.15976","version":1},"attestation_state":"computed","paper":{"title":"Reference-Free Reinforcement Learning Fine-Tuning for MT: A Seq2Seq Perspective","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Carlos Escolano, Ernesto Garcia-Estrada, Jos\\'e A. R. Fonallosa","submitted_at":"2026-05-15T14:11:23Z","abstract_excerpt":"Production machine translation relies overwhelmingly on encoder-decoder Seq2Seq models, yet reinforcement learning approaches to MT fine-tuning have largely targeted decoder-only LLMs at $\\geq$7B parameters, with limited systematic study of encoder-decoder architectures. We apply Group Relative Policy Optimization to NLLB-200 (600M and 1.3B) using a hybrid reference-free reward (LaBSE and COMET-Kiwi) that requires no parallel data at fine-tuning time, evaluating across 13 typologically diverse languages. GRPO yields consistent improvements on all 13 languages, up to $+$5.03 chrF++ for Traditio"},"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":"2605.15976","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-15T14:11:23Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"39c6359d0f8c9dadc27ce9056d049af581a6414c86214c834116b86fd0ad1d23","abstract_canon_sha256":"2f77d85544eec36d3d6d2780d2eb5b09b5f657e9e77436aee8ecce54c9e8fddf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:01:47.339264Z","signature_b64":"Nyb+FaZEpywBAFb9FAozsZioP12uJI8qf2mu67rA4qg7yL9EMmEyktn4+im9KEO+lZaVXeZknbCiv5BFCqJeCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"88a099a48e9edc016a95d6171d252fa7e4b598faeb2847d8bc881e561c9b1538","last_reissued_at":"2026-05-20T00:01:47.338652Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:01:47.338652Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reference-Free Reinforcement Learning Fine-Tuning for MT: A Seq2Seq Perspective","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Carlos Escolano, Ernesto Garcia-Estrada, Jos\\'e A. R. Fonallosa","submitted_at":"2026-05-15T14:11:23Z","abstract_excerpt":"Production machine translation relies overwhelmingly on encoder-decoder Seq2Seq models, yet reinforcement learning approaches to MT fine-tuning have largely targeted decoder-only LLMs at $\\geq$7B parameters, with limited systematic study of encoder-decoder architectures. We apply Group Relative Policy Optimization to NLLB-200 (600M and 1.3B) using a hybrid reference-free reward (LaBSE and COMET-Kiwi) that requires no parallel data at fine-tuning time, evaluating across 13 typologically diverse languages. GRPO yields consistent improvements on all 13 languages, up to $+$5.03 chrF++ for Traditio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.15976","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/2605.15976/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-19T17:33:44.864180Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T17:01:55.683966Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"b7ba2e7b7a19dbeacc06af55080bfb9d356b99332f1a04a1ca8137255a560695"},"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":"2605.15976","created_at":"2026-05-20T00:01:47.338736+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15976v1","created_at":"2026-05-20T00:01:47.338736+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15976","created_at":"2026-05-20T00:01:47.338736+00:00"},{"alias_kind":"pith_short_12","alias_value":"RCQJTJEOT3OA","created_at":"2026-05-20T00:01:47.338736+00:00"},{"alias_kind":"pith_short_16","alias_value":"RCQJTJEOT3OAC2UV","created_at":"2026-05-20T00:01:47.338736+00:00"},{"alias_kind":"pith_short_8","alias_value":"RCQJTJEO","created_at":"2026-05-20T00:01:47.338736+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/RCQJTJEOT3OAC2UV2YLR2JJPU7","json":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7.json","graph_json":"https://pith.science/api/pith-number/RCQJTJEOT3OAC2UV2YLR2JJPU7/graph.json","events_json":"https://pith.science/api/pith-number/RCQJTJEOT3OAC2UV2YLR2JJPU7/events.json","paper":"https://pith.science/paper/RCQJTJEO"},"agent_actions":{"view_html":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7","download_json":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7.json","view_paper":"https://pith.science/paper/RCQJTJEO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15976&json=true","fetch_graph":"https://pith.science/api/pith-number/RCQJTJEOT3OAC2UV2YLR2JJPU7/graph.json","fetch_events":"https://pith.science/api/pith-number/RCQJTJEOT3OAC2UV2YLR2JJPU7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7/action/storage_attestation","attest_author":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7/action/author_attestation","sign_citation":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7/action/citation_signature","submit_replication":"https://pith.science/pith/RCQJTJEOT3OAC2UV2YLR2JJPU7/action/replication_record"}},"created_at":"2026-05-20T00:01:47.338736+00:00","updated_at":"2026-05-20T00:01:47.338736+00:00"}