{"paper":{"title":"Hierarchical Long-Term Semantic Memory for LinkedIn's Hiring Agent","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"A schema-aligned hierarchical memory tree lets LLM agents store and retrieve long-term semantic knowledge with over 10% gains in correctness and retrieval quality.","cross_cats":["cs.LG"],"primary_cat":"cs.IR","authors_text":"Emir Poyraz, Karthik Ramgopal, Praveen Kumar Bodigutla, Shangjin Zhang, Xiaofeng Wang, Xiaoyang Gu, Xie Lu, Ye Jin, Yvonne Li, Zhentao Xu","submitted_at":"2026-04-29T00:53:52Z","abstract_excerpt":"Large Language Model (LLM) agents are increasingly used in real-world products, where personalized and context-aware user interactions are essential. A central enabler of such capabilities is the agent's long-term semantic memory system, which extracts implicit and explicit signals from noisy longitudinal behavioral data, stores them in a structured form, and supports low-latency retrieval. Building industrial-grade long-term memory for LLM agents raises five challenges: scalability, low-latency retrieval, privacy constraints, adaptability, and observability. We introduce the Hierarchical Long"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"HLTM improves answer correctness and retrieval F1 significantly by more than 10%, while significantly advancing the Pareto frontier between query and indexing latency. HLTM has been deployed in LinkedIn's Hiring Assistant to power core personalization features in production hiring workflows.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the schema-aligned memory tree and adaptation mechanism generalize across diverse use cases and that the claimed >10% gains on LinkedIn's internal Hiring Assistant data reflect real-world improvements without undisclosed data selection or baseline choices.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"HLTM builds a hierarchical memory tree from longitudinal data to enable scalable, private, low-latency retrieval, delivering over 10% gains in answer correctness and retrieval F1 for LinkedIn's Hiring Assistant while improving the query-indexing latency tradeoff.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A schema-aligned hierarchical memory tree lets LLM agents store and retrieve long-term semantic knowledge with over 10% gains in correctness and retrieval quality.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"460d19f0b33f857ea61f3cce5e7a32ad7f632c32ff15b2607210da7f7342729f"},"source":{"id":"2604.26197","kind":"arxiv","version":2},"verdict":{"id":"55b68b54-38f8-430b-a5cd-1540be7b5557","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-07T13:28:13.899470Z","strongest_claim":"HLTM improves answer correctness and retrieval F1 significantly by more than 10%, while significantly advancing the Pareto frontier between query and indexing latency. HLTM has been deployed in LinkedIn's Hiring Assistant to power core personalization features in production hiring workflows.","one_line_summary":"HLTM builds a hierarchical memory tree from longitudinal data to enable scalable, private, low-latency retrieval, delivering over 10% gains in answer correctness and retrieval F1 for LinkedIn's Hiring Assistant while improving the query-indexing latency tradeoff.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the schema-aligned memory tree and adaptation mechanism generalize across diverse use cases and that the claimed >10% gains on LinkedIn's internal Hiring Assistant data reflect real-world improvements without undisclosed data selection or baseline choices.","pith_extraction_headline":"A schema-aligned hierarchical memory tree lets LLM agents store and retrieve long-term semantic knowledge with over 10% gains in correctness and retrieval quality."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.26197/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T00:40:45.255151Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T20:24:47.788660Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"3fc9a170e6d1ee9608880e0960a4e1e247c83b22ee7ade4667f9245ebab3e137"},"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"}