{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QCGQRBGPSFSKF3RGSCIWYYFR7L","short_pith_number":"pith:QCGQRBGP","schema_version":"1.0","canonical_sha256":"808d0884cf9164a2ee2690916c60b1faf93478b5abd48082a61c891790ebf2b4","source":{"kind":"arxiv","id":"2411.03524","version":1},"attestation_state":"computed","paper":{"title":"Mitigating Metric Bias in Minimum Bayes Risk Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Daniel Deutsch, Geza Kovacs, Markus Freitag","submitted_at":"2024-11-05T22:01:27Z","abstract_excerpt":"While Minimum Bayes Risk (MBR) decoding using metrics such as COMET or MetricX has outperformed traditional decoding methods such as greedy or beam search, it introduces a challenge we refer to as metric bias. As MBR decoding aims to produce translations that score highly according to a specific utility metric, this very process makes it impossible to use the same metric for both decoding and evaluation, as improvements might simply be due to reward hacking rather than reflecting real quality improvements. In this work we find that compared to human ratings, neural metrics not only overestimat"},"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":"2411.03524","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-05T22:01:27Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0265b720cbab60374c94033691d8645d5f34f47e0c5f2eec8ff683be95d96818","abstract_canon_sha256":"0c44a501d04b2d01bb093b47080af3df210aedb227ffbb9f290fd2ce77331d83"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:31:33.845306Z","signature_b64":"OVGEAmpWmFQfogZ1HmCqOXRLOcsO32gsNaO6VuGwyH858ocEncPonadj+f9LFAfbAhnOjruzCFuH9B9UrXVHCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"808d0884cf9164a2ee2690916c60b1faf93478b5abd48082a61c891790ebf2b4","last_reissued_at":"2026-07-05T09:31:33.844033Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:31:33.844033Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating Metric Bias in Minimum Bayes Risk Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Daniel Deutsch, Geza Kovacs, Markus Freitag","submitted_at":"2024-11-05T22:01:27Z","abstract_excerpt":"While Minimum Bayes Risk (MBR) decoding using metrics such as COMET or MetricX has outperformed traditional decoding methods such as greedy or beam search, it introduces a challenge we refer to as metric bias. As MBR decoding aims to produce translations that score highly according to a specific utility metric, this very process makes it impossible to use the same metric for both decoding and evaluation, as improvements might simply be due to reward hacking rather than reflecting real quality improvements. In this work we find that compared to human ratings, neural metrics not only overestimat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.03524","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/2411.03524/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":"2411.03524","created_at":"2026-07-05T09:31:33.844858+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.03524v1","created_at":"2026-07-05T09:31:33.844858+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.03524","created_at":"2026-07-05T09:31:33.844858+00:00"},{"alias_kind":"pith_short_12","alias_value":"QCGQRBGPSFSK","created_at":"2026-07-05T09:31:33.844858+00:00"},{"alias_kind":"pith_short_16","alias_value":"QCGQRBGPSFSKF3RG","created_at":"2026-07-05T09:31:33.844858+00:00"},{"alias_kind":"pith_short_8","alias_value":"QCGQRBGP","created_at":"2026-07-05T09:31:33.844858+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/QCGQRBGPSFSKF3RGSCIWYYFR7L","json":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L.json","graph_json":"https://pith.science/api/pith-number/QCGQRBGPSFSKF3RGSCIWYYFR7L/graph.json","events_json":"https://pith.science/api/pith-number/QCGQRBGPSFSKF3RGSCIWYYFR7L/events.json","paper":"https://pith.science/paper/QCGQRBGP"},"agent_actions":{"view_html":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L","download_json":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L.json","view_paper":"https://pith.science/paper/QCGQRBGP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.03524&json=true","fetch_graph":"https://pith.science/api/pith-number/QCGQRBGPSFSKF3RGSCIWYYFR7L/graph.json","fetch_events":"https://pith.science/api/pith-number/QCGQRBGPSFSKF3RGSCIWYYFR7L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L/action/storage_attestation","attest_author":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L/action/author_attestation","sign_citation":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L/action/citation_signature","submit_replication":"https://pith.science/pith/QCGQRBGPSFSKF3RGSCIWYYFR7L/action/replication_record"}},"created_at":"2026-07-05T09:31:33.844858+00:00","updated_at":"2026-07-05T09:31:33.844858+00:00"}