{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KL4Z3ILRS3ISPZ27KANJYWVMYA","short_pith_number":"pith:KL4Z3ILR","schema_version":"1.0","canonical_sha256":"52f99da17196d127e75f501a9c5aacc0053361c5db972dc99d460431842e5c0b","source":{"kind":"arxiv","id":"2605.23572","version":1},"attestation_state":"computed","paper":{"title":"HARNESS-LM: A Three-Phase Training Recipe for Harnessing SLMs in Sponsored Search Retrieval","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.IR","authors_text":"Amit Singh, Lakshya Kumar, Manik Varma, Nikit Begwani, Pranjal Chitale, Shikhar Mohan, Vipul Gupta","submitted_at":"2026-05-22T12:39:56Z","abstract_excerpt":"In the competitive landscape of sponsored search, balancing retrieval quality with production latency is a critical challenge. While large retrieval models based on Small Language Models (SLMs) such as Qwen3-Embedding-4B/8B set strong upper bounds on public benchmarks, their deployment in high-throughput, latency-sensitive environments remains impractical. In this paper, we present HARNESS-LM (HLM), a three-phase training framework for transferring the capabilities of large-scale retrievers into compact, cost-efficient models. The approach comprises: (1) training a high-performance reference ("},"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":"2605.23572","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.IR","submitted_at":"2026-05-22T12:39:56Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"6203a436e31c10c872f74eb1f0f026df3af715c50f47e85f225a0ac71817d0b1","abstract_canon_sha256":"6edde732d4ba523cdc57cc624fbbb9387f6081844c470afd6fa96b43e36c8336"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:02:20.184372Z","signature_b64":"+vkJvf7WKYmUiTZXY7LVc9w/db9WnRynuyLa4kN96q6XtRGeq5SbMjQ+tr357EbOKL8ECwPImS1yr+dqLeT/Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"52f99da17196d127e75f501a9c5aacc0053361c5db972dc99d460431842e5c0b","last_reissued_at":"2026-05-25T02:02:20.183387Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:02:20.183387Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HARNESS-LM: A Three-Phase Training Recipe for Harnessing SLMs in Sponsored Search Retrieval","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.IR","authors_text":"Amit Singh, Lakshya Kumar, Manik Varma, Nikit Begwani, Pranjal Chitale, Shikhar Mohan, Vipul Gupta","submitted_at":"2026-05-22T12:39:56Z","abstract_excerpt":"In the competitive landscape of sponsored search, balancing retrieval quality with production latency is a critical challenge. While large retrieval models based on Small Language Models (SLMs) such as Qwen3-Embedding-4B/8B set strong upper bounds on public benchmarks, their deployment in high-throughput, latency-sensitive environments remains impractical. In this paper, we present HARNESS-LM (HLM), a three-phase training framework for transferring the capabilities of large-scale retrievers into compact, cost-efficient models. The approach comprises: (1) training a high-performance reference ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.23572","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/2605.23572/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":"2605.23572","created_at":"2026-05-25T02:02:20.183524+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.23572v1","created_at":"2026-05-25T02:02:20.183524+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.23572","created_at":"2026-05-25T02:02:20.183524+00:00"},{"alias_kind":"pith_short_12","alias_value":"KL4Z3ILRS3IS","created_at":"2026-05-25T02:02:20.183524+00:00"},{"alias_kind":"pith_short_16","alias_value":"KL4Z3ILRS3ISPZ27","created_at":"2026-05-25T02:02:20.183524+00:00"},{"alias_kind":"pith_short_8","alias_value":"KL4Z3ILR","created_at":"2026-05-25T02:02:20.183524+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/KL4Z3ILRS3ISPZ27KANJYWVMYA","json":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA.json","graph_json":"https://pith.science/api/pith-number/KL4Z3ILRS3ISPZ27KANJYWVMYA/graph.json","events_json":"https://pith.science/api/pith-number/KL4Z3ILRS3ISPZ27KANJYWVMYA/events.json","paper":"https://pith.science/paper/KL4Z3ILR"},"agent_actions":{"view_html":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA","download_json":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA.json","view_paper":"https://pith.science/paper/KL4Z3ILR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.23572&json=true","fetch_graph":"https://pith.science/api/pith-number/KL4Z3ILRS3ISPZ27KANJYWVMYA/graph.json","fetch_events":"https://pith.science/api/pith-number/KL4Z3ILRS3ISPZ27KANJYWVMYA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA/action/storage_attestation","attest_author":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA/action/author_attestation","sign_citation":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA/action/citation_signature","submit_replication":"https://pith.science/pith/KL4Z3ILRS3ISPZ27KANJYWVMYA/action/replication_record"}},"created_at":"2026-05-25T02:02:20.183524+00:00","updated_at":"2026-05-25T02:02:20.183524+00:00"}