{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:7IZAURTHYJN43EG2DM6QYRJUHA","short_pith_number":"pith:7IZAURTH","schema_version":"1.0","canonical_sha256":"fa320a4667c25bcd90da1b3d0c45343826ca12389439f2a6f1a8cfcb51feb7b7","source":{"kind":"arxiv","id":"1805.10627","version":3},"attestation_state":"computed","paper":{"title":"Reliability and Learnability of Human Bandit Feedback for Sequence-to-Sequence Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.CL","authors_text":"Joshua Uyheng, Julia Kreutzer, Stefan Riezler","submitted_at":"2018-05-27T13:51:48Z","abstract_excerpt":"We present a study on reinforcement learning (RL) from human bandit feedback for sequence-to-sequence learning, exemplified by the task of bandit neural machine translation (NMT). We investigate the reliability of human bandit feedback, and analyze the influence of reliability on the learnability of a reward estimator, and the effect of the quality of reward estimates on the overall RL task. Our analysis of cardinal (5-point ratings) and ordinal (pairwise preferences) feedback shows that their intra- and inter-annotator $\\alpha$-agreement is comparable. Best reliability is obtained for standar"},"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":"1805.10627","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2018-05-27T13:51:48Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"2254f5dc1ddaa8a3df117fd52039c5ee83480555a0117dfa4fd9da7a035e27d7","abstract_canon_sha256":"be66dd2d0000906a95913f1ea58177c5bc6c12b316d5b677ba9c06e8957cf3ae"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:58:22.995105Z","signature_b64":"Qxr6/cAIE4TXsOUB8HJ5i2GgwqtjLsomX237BA1lgYpwfaBN3dxQHu1U5qh3HYqRJSXGxGtZoGWyFkHAaCr4DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fa320a4667c25bcd90da1b3d0c45343826ca12389439f2a6f1a8cfcb51feb7b7","last_reissued_at":"2026-05-17T23:58:22.994297Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:58:22.994297Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reliability and Learnability of Human Bandit Feedback for Sequence-to-Sequence Reinforcement Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.CL","authors_text":"Joshua Uyheng, Julia Kreutzer, Stefan Riezler","submitted_at":"2018-05-27T13:51:48Z","abstract_excerpt":"We present a study on reinforcement learning (RL) from human bandit feedback for sequence-to-sequence learning, exemplified by the task of bandit neural machine translation (NMT). We investigate the reliability of human bandit feedback, and analyze the influence of reliability on the learnability of a reward estimator, and the effect of the quality of reward estimates on the overall RL task. Our analysis of cardinal (5-point ratings) and ordinal (pairwise preferences) feedback shows that their intra- and inter-annotator $\\alpha$-agreement is comparable. Best reliability is obtained for standar"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1805.10627","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"1805.10627","created_at":"2026-05-17T23:58:22.994445+00:00"},{"alias_kind":"arxiv_version","alias_value":"1805.10627v3","created_at":"2026-05-17T23:58:22.994445+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1805.10627","created_at":"2026-05-17T23:58:22.994445+00:00"},{"alias_kind":"pith_short_12","alias_value":"7IZAURTHYJN4","created_at":"2026-05-18T12:32:11.075285+00:00"},{"alias_kind":"pith_short_16","alias_value":"7IZAURTHYJN43EG2","created_at":"2026-05-18T12:32:11.075285+00:00"},{"alias_kind":"pith_short_8","alias_value":"7IZAURTH","created_at":"2026-05-18T12:32:11.075285+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2402.13228","citing_title":"Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2409.12917","citing_title":"Training Language Models to Self-Correct via Reinforcement Learning","ref_index":172,"is_internal_anchor":true},{"citing_arxiv_id":"1909.08593","citing_title":"Fine-Tuning Language Models from Human Preferences","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA","json":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA.json","graph_json":"https://pith.science/api/pith-number/7IZAURTHYJN43EG2DM6QYRJUHA/graph.json","events_json":"https://pith.science/api/pith-number/7IZAURTHYJN43EG2DM6QYRJUHA/events.json","paper":"https://pith.science/paper/7IZAURTH"},"agent_actions":{"view_html":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA","download_json":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA.json","view_paper":"https://pith.science/paper/7IZAURTH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1805.10627&json=true","fetch_graph":"https://pith.science/api/pith-number/7IZAURTHYJN43EG2DM6QYRJUHA/graph.json","fetch_events":"https://pith.science/api/pith-number/7IZAURTHYJN43EG2DM6QYRJUHA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA/action/storage_attestation","attest_author":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA/action/author_attestation","sign_citation":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA/action/citation_signature","submit_replication":"https://pith.science/pith/7IZAURTHYJN43EG2DM6QYRJUHA/action/replication_record"}},"created_at":"2026-05-17T23:58:22.994445+00:00","updated_at":"2026-05-17T23:58:22.994445+00:00"}