{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:U4OWUOB73OG6PYPAOZZJZ7QHD3","short_pith_number":"pith:U4OWUOB7","schema_version":"1.0","canonical_sha256":"a71d6a383fdb8de7e1e076729cfe071eff41e61b1d0ea079170403cf1f798677","source":{"kind":"arxiv","id":"2605.26385","version":1},"attestation_state":"computed","paper":{"title":"Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.IR","authors_text":"Ana-Roxana Pop, Ariel Evnine, Haruka Kiyohara, Israel Nir, Mihaela Curmei, Nitzan Razin, Sarah Dean, Shankar Kalyanaraman, Thorsten Joachims, Udi Weinsberg","submitted_at":"2026-05-25T23:17:37Z","abstract_excerpt":"Large-scale search, recommendation, and retrieval-augmented generation (RAG) systems typically employ a two-stage architecture: an early-stage ranker (ESR) generates a candidate set, which is subsequently re-ranked by a late-stage ranker (LSR). While there are many reinforcement learning (RL) methods for training the LSR, end-to-end training of the ESR has proven challenging. In particular, naive application of \"vanilla\" policy gradient (V-PG) is not scalable for candidate-set sizes relevant for practical use due to exploding variance. This issue arises because V-PG propagates the gradient to "},"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":"2605.26385","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.IR","submitted_at":"2026-05-25T23:17:37Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"33a3be9329d7bdca8fe0c8318e88d207f9f41ff374e50b4ea34706df0e92d725","abstract_canon_sha256":"36b073b4f41ab8cab6d2bb6d0daac8658756d44259fd59163e4dd9b91a9048b2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:05:15.397817Z","signature_b64":"CSjqDmbiryapVjbE3UurqaEmLDU9hEMidUOhYFVfkuI1BSxKTDAetEZb2L2sCuMPA2jqyk0fTnMgSdLxfJJhCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a71d6a383fdb8de7e1e076729cfe071eff41e61b1d0ea079170403cf1f798677","last_reissued_at":"2026-05-27T01:05:15.397023Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:05:15.397023Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Credit-assigned Policy Gradient for Early Stage Retrieval in Two-stage Ranking","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.IR","authors_text":"Ana-Roxana Pop, Ariel Evnine, Haruka Kiyohara, Israel Nir, Mihaela Curmei, Nitzan Razin, Sarah Dean, Shankar Kalyanaraman, Thorsten Joachims, Udi Weinsberg","submitted_at":"2026-05-25T23:17:37Z","abstract_excerpt":"Large-scale search, recommendation, and retrieval-augmented generation (RAG) systems typically employ a two-stage architecture: an early-stage ranker (ESR) generates a candidate set, which is subsequently re-ranked by a late-stage ranker (LSR). While there are many reinforcement learning (RL) methods for training the LSR, end-to-end training of the ESR has proven challenging. In particular, naive application of \"vanilla\" policy gradient (V-PG) is not scalable for candidate-set sizes relevant for practical use due to exploding variance. This issue arises because V-PG propagates the gradient to "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26385","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/2605.26385/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":"2605.26385","created_at":"2026-05-27T01:05:15.397155+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.26385v1","created_at":"2026-05-27T01:05:15.397155+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26385","created_at":"2026-05-27T01:05:15.397155+00:00"},{"alias_kind":"pith_short_12","alias_value":"U4OWUOB73OG6","created_at":"2026-05-27T01:05:15.397155+00:00"},{"alias_kind":"pith_short_16","alias_value":"U4OWUOB73OG6PYPA","created_at":"2026-05-27T01:05:15.397155+00:00"},{"alias_kind":"pith_short_8","alias_value":"U4OWUOB7","created_at":"2026-05-27T01:05:15.397155+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/U4OWUOB73OG6PYPAOZZJZ7QHD3","json":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3.json","graph_json":"https://pith.science/api/pith-number/U4OWUOB73OG6PYPAOZZJZ7QHD3/graph.json","events_json":"https://pith.science/api/pith-number/U4OWUOB73OG6PYPAOZZJZ7QHD3/events.json","paper":"https://pith.science/paper/U4OWUOB7"},"agent_actions":{"view_html":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3","download_json":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3.json","view_paper":"https://pith.science/paper/U4OWUOB7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.26385&json=true","fetch_graph":"https://pith.science/api/pith-number/U4OWUOB73OG6PYPAOZZJZ7QHD3/graph.json","fetch_events":"https://pith.science/api/pith-number/U4OWUOB73OG6PYPAOZZJZ7QHD3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3/action/storage_attestation","attest_author":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3/action/author_attestation","sign_citation":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3/action/citation_signature","submit_replication":"https://pith.science/pith/U4OWUOB73OG6PYPAOZZJZ7QHD3/action/replication_record"}},"created_at":"2026-05-27T01:05:15.397155+00:00","updated_at":"2026-05-27T01:05:15.397155+00:00"}