{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XIJWZUHOVRAMXG4HYE3FOWWHWD","short_pith_number":"pith:XIJWZUHO","schema_version":"1.0","canonical_sha256":"ba136cd0eeac40cb9b87c136575ac7b0eda29eee296e6b2097e7325ad0324541","source":{"kind":"arxiv","id":"2602.03352","version":2},"attestation_state":"computed","paper":{"title":"PEGRL: Improving Machine Translation by Post-Editing Guided Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hao Zhou, Junlan Feng, Shujian Huang, Xin Huang, Xue Han, Yunzhi Shen","submitted_at":"2026-02-03T10:22:55Z","abstract_excerpt":"Reinforcement learning (RL) has shown strong promise for LLM-based machine translation, with recent methods such as GRPO demonstrating notable gains; nevertheless, translation-oriented RL remains challenged by noisy learning signals arising from Monte Carlo return estimation, as well as a large trajectory space that favors global exploration over fine-grained local optimization. We introduce \\textbf{PEGRL}, a \\textit{two-stage} RL framework that uses post-editing as an auxiliary task to stabilize training and guide overall optimization. At each iteration, translation outputs are sampled to con"},"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":"2602.03352","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-02-03T10:22:55Z","cross_cats_sorted":[],"title_canon_sha256":"c46dab859a6e8d16ed18f2e3a34a9c787cbd2d950c55f5e951a9971037142a5d","abstract_canon_sha256":"7cee53c45dbb9d6ae6d8bf1545b0a1c0142dd80dfc8650ce461f9214283f2c7a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:24.119844Z","signature_b64":"eahCLaNuKUjAdU8q1rytVOauHHkhAM1rnQYoj6UlntY7xo/6d2awNrHTKC8EdyffBbGbensfElvaZi3MIds3CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ba136cd0eeac40cb9b87c136575ac7b0eda29eee296e6b2097e7325ad0324541","last_reissued_at":"2026-05-20T00:04:24.119095Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:24.119095Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PEGRL: Improving Machine Translation by Post-Editing Guided Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hao Zhou, Junlan Feng, Shujian Huang, Xin Huang, Xue Han, Yunzhi Shen","submitted_at":"2026-02-03T10:22:55Z","abstract_excerpt":"Reinforcement learning (RL) has shown strong promise for LLM-based machine translation, with recent methods such as GRPO demonstrating notable gains; nevertheless, translation-oriented RL remains challenged by noisy learning signals arising from Monte Carlo return estimation, as well as a large trajectory space that favors global exploration over fine-grained local optimization. We introduce \\textbf{PEGRL}, a \\textit{two-stage} RL framework that uses post-editing as an auxiliary task to stabilize training and guide overall optimization. At each iteration, translation outputs are sampled to con"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.03352","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/2602.03352/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":"2602.03352","created_at":"2026-05-20T00:04:24.119215+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.03352v2","created_at":"2026-05-20T00:04:24.119215+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.03352","created_at":"2026-05-20T00:04:24.119215+00:00"},{"alias_kind":"pith_short_12","alias_value":"XIJWZUHOVRAM","created_at":"2026-05-20T00:04:24.119215+00:00"},{"alias_kind":"pith_short_16","alias_value":"XIJWZUHOVRAMXG4H","created_at":"2026-05-20T00:04:24.119215+00:00"},{"alias_kind":"pith_short_8","alias_value":"XIJWZUHO","created_at":"2026-05-20T00:04:24.119215+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.25702","citing_title":"Backtranslation Augmented Direct Preference Optimization for Neural Machine Translation","ref_index":9,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD","json":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD.json","graph_json":"https://pith.science/api/pith-number/XIJWZUHOVRAMXG4HYE3FOWWHWD/graph.json","events_json":"https://pith.science/api/pith-number/XIJWZUHOVRAMXG4HYE3FOWWHWD/events.json","paper":"https://pith.science/paper/XIJWZUHO"},"agent_actions":{"view_html":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD","download_json":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD.json","view_paper":"https://pith.science/paper/XIJWZUHO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.03352&json=true","fetch_graph":"https://pith.science/api/pith-number/XIJWZUHOVRAMXG4HYE3FOWWHWD/graph.json","fetch_events":"https://pith.science/api/pith-number/XIJWZUHOVRAMXG4HYE3FOWWHWD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD/action/storage_attestation","attest_author":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD/action/author_attestation","sign_citation":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD/action/citation_signature","submit_replication":"https://pith.science/pith/XIJWZUHOVRAMXG4HYE3FOWWHWD/action/replication_record"}},"created_at":"2026-05-20T00:04:24.119215+00:00","updated_at":"2026-05-20T00:04:24.119215+00:00"}