{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FBBRPPSJ2AVO4FMJTQMNON5BRQ","short_pith_number":"pith:FBBRPPSJ","schema_version":"1.0","canonical_sha256":"284317be49d02aee15899c18d737a18c00e664b86fb1dc3bde671ff8b22a8d2a","source":{"kind":"arxiv","id":"2504.13630","version":1},"attestation_state":"computed","paper":{"title":"Remedy: Learning Machine Translation Evaluation from Human Preferences with Reward Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christof Monz, Shaomu Tan","submitted_at":"2025-04-18T11:11:14Z","abstract_excerpt":"A key challenge in MT evaluation is the inherent noise and inconsistency of human ratings. Regression-based neural metrics struggle with this noise, while prompting LLMs shows promise at system-level evaluation but performs poorly at segment level. In this work, we propose ReMedy, a novel MT metric framework that reformulates translation evaluation as a reward modeling task. Instead of regressing on imperfect human ratings directly, ReMedy learns relative translation quality using pairwise preference data, resulting in a more reliable evaluation. In extensive experiments across WMT22-24 shared"},"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":"2504.13630","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-18T11:11:14Z","cross_cats_sorted":[],"title_canon_sha256":"fafdbd8eb82d028c14928abecaec26c56e57b46ea53239ae9b2ca7f7a9fd634c","abstract_canon_sha256":"2ca4ad7ec94dbe7b7a75ba58f2b344e7423e331813bc444235cea7372e7d2947"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:50:59.360506Z","signature_b64":"ezDuUzgLnfM64cmnoHc72Jv2eHsEbVhfYt3lW9JSMYC7FgCKLt9f6QpILqo4JnAV5p9d8E95VaAJqUb1deTNDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"284317be49d02aee15899c18d737a18c00e664b86fb1dc3bde671ff8b22a8d2a","last_reissued_at":"2026-07-05T10:50:59.360033Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:50:59.360033Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Remedy: Learning Machine Translation Evaluation from Human Preferences with Reward Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christof Monz, Shaomu Tan","submitted_at":"2025-04-18T11:11:14Z","abstract_excerpt":"A key challenge in MT evaluation is the inherent noise and inconsistency of human ratings. Regression-based neural metrics struggle with this noise, while prompting LLMs shows promise at system-level evaluation but performs poorly at segment level. In this work, we propose ReMedy, a novel MT metric framework that reformulates translation evaluation as a reward modeling task. Instead of regressing on imperfect human ratings directly, ReMedy learns relative translation quality using pairwise preference data, resulting in a more reliable evaluation. In extensive experiments across WMT22-24 shared"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.13630","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/2504.13630/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":"2504.13630","created_at":"2026-07-05T10:50:59.360089+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.13630v1","created_at":"2026-07-05T10:50:59.360089+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.13630","created_at":"2026-07-05T10:50:59.360089+00:00"},{"alias_kind":"pith_short_12","alias_value":"FBBRPPSJ2AVO","created_at":"2026-07-05T10:50:59.360089+00:00"},{"alias_kind":"pith_short_16","alias_value":"FBBRPPSJ2AVO4FMJ","created_at":"2026-07-05T10:50:59.360089+00:00"},{"alias_kind":"pith_short_8","alias_value":"FBBRPPSJ","created_at":"2026-07-05T10:50:59.360089+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.17393","citing_title":"Who Watches the Watchmen? Humans Disagree With Translation Metrics on Unseen Domains","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ","json":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ.json","graph_json":"https://pith.science/api/pith-number/FBBRPPSJ2AVO4FMJTQMNON5BRQ/graph.json","events_json":"https://pith.science/api/pith-number/FBBRPPSJ2AVO4FMJTQMNON5BRQ/events.json","paper":"https://pith.science/paper/FBBRPPSJ"},"agent_actions":{"view_html":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ","download_json":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ.json","view_paper":"https://pith.science/paper/FBBRPPSJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.13630&json=true","fetch_graph":"https://pith.science/api/pith-number/FBBRPPSJ2AVO4FMJTQMNON5BRQ/graph.json","fetch_events":"https://pith.science/api/pith-number/FBBRPPSJ2AVO4FMJTQMNON5BRQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ/action/storage_attestation","attest_author":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ/action/author_attestation","sign_citation":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ/action/citation_signature","submit_replication":"https://pith.science/pith/FBBRPPSJ2AVO4FMJTQMNON5BRQ/action/replication_record"}},"created_at":"2026-07-05T10:50:59.360089+00:00","updated_at":"2026-07-05T10:50:59.360089+00:00"}