{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UFXHD4HBH6SNB3LUIE7MYNLNTD","short_pith_number":"pith:UFXHD4HB","schema_version":"1.0","canonical_sha256":"a16e71f0e13fa4d0ed74413ecc356d98ee92d2d3438853165af46a4321ead46e","source":{"kind":"arxiv","id":"2605.14434","version":1},"attestation_state":"computed","paper":{"title":"Efficient Generative Retrieval for E-commerce Search with Semantic Cluster IDs and Expert-Guided RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Category-and-query constrained semantic IDs enable generative retrieval as a practical recall supplement in e-commerce search by halving beam search size and lifting click hit rates.","cross_cats":["cs.AI"],"primary_cat":"cs.IR","authors_text":"Bokang Wang, Guangxin Song, Jianbo Zhu, Jing Wang, Junjie Bai, Mingmin Jin, Xing Fang, Zhenyu Xie","submitted_at":"2026-05-14T06:27:46Z","abstract_excerpt":"Generative retrieval offers a promising alternative by unifying the fragmented multi-stage retrieval process into a single end-to-end model. However, its practical adoption in industrial e-commerce search remains challenging, given the massive and dynamic product catalogs, strict latency requirements, and the need to align retrieval with downstream ranking goals. In this work, we propose a retrieval framework tailored for real-world recall scenarios, positioning generative retrieval as a recall-stage supplement rather than an end-to-end replacement. Our method, CQ-SID (Category-and-Query const"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.14434","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.IR","submitted_at":"2026-05-14T06:27:46Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"8626fc65b8ab07af7b888edf88145bd6b7ebe3f6d874179a20e61110cdadb773","abstract_canon_sha256":"3c8188886fb77ba37a1c969c50887122c397947bb9f59318a84cfc8a5ef91053"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:07.097782Z","signature_b64":"qpdVog4s9JTRoGrqeQwhjoHAzLDouCebdVdKHScbhnXDPA/DD22Tw/3TUIm6u9BkHpDsMp5UiZprq0JkGiB+Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a16e71f0e13fa4d0ed74413ecc356d98ee92d2d3438853165af46a4321ead46e","last_reissued_at":"2026-05-17T23:39:07.097082Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:07.097082Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Generative Retrieval for E-commerce Search with Semantic Cluster IDs and Expert-Guided RL","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Category-and-query constrained semantic IDs enable generative retrieval as a practical recall supplement in e-commerce search by halving beam search size and lifting click hit rates.","cross_cats":["cs.AI"],"primary_cat":"cs.IR","authors_text":"Bokang Wang, Guangxin Song, Jianbo Zhu, Jing Wang, Junjie Bai, Mingmin Jin, Xing Fang, Zhenyu Xie","submitted_at":"2026-05-14T06:27:46Z","abstract_excerpt":"Generative retrieval offers a promising alternative by unifying the fragmented multi-stage retrieval process into a single end-to-end model. However, its practical adoption in industrial e-commerce search remains challenging, given the massive and dynamic product catalogs, strict latency requirements, and the need to align retrieval with downstream ranking goals. In this work, we propose a retrieval framework tailored for real-world recall scenarios, positioning generative retrieval as a recall-stage supplement rather than an end-to-end replacement. Our method, CQ-SID (Category-and-Query const"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"CQ-SID achieves up to 26.76% and 11.11% relative gains in semantic and personalized click hitrate over RQ-VAE baselines, while halving beam search size. EG-GRPO further improves multi-objective performance. Online A/B tests confirm gains in GMV (+1.15%) and UCTCVR (+0.40%). The generative recall channel now contributes substantially in production, accounting for over 50.25% of exposures, 58.96% of clicks, and 72.63% of purchases.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the hierarchical semantic cluster IDs produced by category-and-query constrained contrastive learning plus Residual Quantized VAEs preserve sufficient relevance signals to support effective beam search and downstream ranking alignment under the sparse-reward RL regime described.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CQ-SID semantic IDs and EG-GRPO RL improve generative retrieval hit rates up to 26.76% over RQ-VAE baselines and deliver +1.15% GMV in live e-commerce A/B tests.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Category-and-query constrained semantic IDs enable generative retrieval as a practical recall supplement in e-commerce search by halving beam search size and lifting click hit rates.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3a57588472926303ad54852cc6a605d4cebbee62d5f8fe5a1b22501cc1f0b85c"},"source":{"id":"2605.14434","kind":"arxiv","version":1},"verdict":{"id":"f3a2e0a4-91b1-4b8c-98f5-3f49185c0af6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T02:12:41.536514Z","strongest_claim":"CQ-SID achieves up to 26.76% and 11.11% relative gains in semantic and personalized click hitrate over RQ-VAE baselines, while halving beam search size. EG-GRPO further improves multi-objective performance. Online A/B tests confirm gains in GMV (+1.15%) and UCTCVR (+0.40%). The generative recall channel now contributes substantially in production, accounting for over 50.25% of exposures, 58.96% of clicks, and 72.63% of purchases.","one_line_summary":"CQ-SID semantic IDs and EG-GRPO RL improve generative retrieval hit rates up to 26.76% over RQ-VAE baselines and deliver +1.15% GMV in live e-commerce A/B tests.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the hierarchical semantic cluster IDs produced by category-and-query constrained contrastive learning plus Residual Quantized VAEs preserve sufficient relevance signals to support effective beam search and downstream ranking alignment under the sparse-reward RL regime described.","pith_extraction_headline":"Category-and-query constrained semantic IDs enable generative retrieval as a practical recall supplement in e-commerce search by halving beam search size and lifting click hit rates."},"references":{"count":26,"sample":[{"doi":"","year":2022,"title":"Michele Bevilacqua, Giuseppe Ottaviano, Patrick Lewis, Scott Yih, Sebastian Riedel, and Fabio Petroni. 2022. Autoregressive search engines: Generating substrings as document identifiers.Advances in Ne","work_id":"4fa971de-b021-47e2-8899-0bca334a154f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"arXiv preprint arXiv:2509.03236 , year=","work_id":"70abc516-1c36-4952-9a48-60b396c6e685","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Jiehan Cheng, Zhicheng Dou, Yutao Zhu, and Xiaoxi Li. 2025. Descriptive and discriminative document identifiers for generative retrieval. InProceedings of the AAAI Conference on Artificial Intelligenc","work_id":"179355f5-c4c5-4bd0-bcf4-5bfa807b4912","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Yingjun Dai and Ahmed El-Roby. 2025. RQ-Rec: Residual Quantized Hierarchical Preference Modeling for Cross-Domain Recommendation. InProceedings of the 33rd ACM International Conference on Multimedia. ","work_id":"e117b143-cd32-4071-a566-10b605bc6359","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"OneRec: Unifying Retrieve and Rank with Generative Recommender and Iterative Preference Alignment","work_id":"d1a07d92-e045-4af2-a79f-c7b0112cf824","ref_index":5,"cited_arxiv_id":"2502.18965","is_internal_anchor":true}],"resolved_work":26,"snapshot_sha256":"2f4a5ebb518cb3e029647cf44d995a3e3d7f56675dbe1e63ffdbc5095b433a20","internal_anchors":5},"formal_canon":{"evidence_count":2,"snapshot_sha256":"4b07650292d7e2e5ce189d3ec997812a43207eef7c0446f70bea36bccf084a87"},"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":"2605.14434","created_at":"2026-05-17T23:39:07.097200+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.14434v1","created_at":"2026-05-17T23:39:07.097200+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14434","created_at":"2026-05-17T23:39:07.097200+00:00"},{"alias_kind":"pith_short_12","alias_value":"UFXHD4HBH6SN","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"UFXHD4HBH6SNB3LU","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"UFXHD4HB","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD","json":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD.json","graph_json":"https://pith.science/api/pith-number/UFXHD4HBH6SNB3LUIE7MYNLNTD/graph.json","events_json":"https://pith.science/api/pith-number/UFXHD4HBH6SNB3LUIE7MYNLNTD/events.json","paper":"https://pith.science/paper/UFXHD4HB"},"agent_actions":{"view_html":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD","download_json":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD.json","view_paper":"https://pith.science/paper/UFXHD4HB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.14434&json=true","fetch_graph":"https://pith.science/api/pith-number/UFXHD4HBH6SNB3LUIE7MYNLNTD/graph.json","fetch_events":"https://pith.science/api/pith-number/UFXHD4HBH6SNB3LUIE7MYNLNTD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD/action/storage_attestation","attest_author":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD/action/author_attestation","sign_citation":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD/action/citation_signature","submit_replication":"https://pith.science/pith/UFXHD4HBH6SNB3LUIE7MYNLNTD/action/replication_record"}},"created_at":"2026-05-17T23:39:07.097200+00:00","updated_at":"2026-05-17T23:39:07.097200+00:00"}