{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NHPKZ2KYHYYS63NWRZE52XKFN4","short_pith_number":"pith:NHPKZ2KY","schema_version":"1.0","canonical_sha256":"69deace9583e312f6db68e49dd5d456f1bf06e5904a82166aafb38cea793362e","source":{"kind":"arxiv","id":"2602.21009","version":2},"attestation_state":"computed","paper":{"title":"HiSAC: Hierarchical Sparse Activation Compression for Ultra-long Sequence Modeling in Recommenders","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.IR","authors_text":"Binbin Cao, Changfa Wu, Daixuan Cheng, Jian Wu, Junyu Bi, Kun Yuan, Shuwen Xiao, Yuning Jiang","submitted_at":"2026-02-24T15:28:58Z","abstract_excerpt":"Modern recommender systems leverage ultra-long user behavior sequences to capture dynamic preferences, but end-to-end modeling is infeasible in production due to latency and memory constraints. While summarizing history via interest centers offers a practical alternative, existing methods struggle to (1) identify user-specific centers at appropriate granularity and (2) accurately assign behaviors, leading to quantization errors and loss of long-tail preferences. To alleviate these issues, we propose Hierarchical Sparse Activation Compression (HiSAC), an efficient framework for personalized seq"},"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":"2602.21009","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.IR","submitted_at":"2026-02-24T15:28:58Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"ceb4e58abae2e55d16600a2cf3735283d2669dd66c9c0e5a4d043e5b42863a39","abstract_canon_sha256":"f0334b2d93144692421e4406a882877a421817a9db6b9b68eaa4b9f4db69149b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T01:16:06.403246Z","signature_b64":"lY97h+tDvFzBh2bvP/yFYwluy61SKhgBmPW1EDGbS6UXPR/+Bl+1DFLWH1whH6eAqgY0bisAL+JYjWZWVoH1Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"69deace9583e312f6db68e49dd5d456f1bf06e5904a82166aafb38cea793362e","last_reissued_at":"2026-07-07T01:16:06.402172Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T01:16:06.402172Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HiSAC: Hierarchical Sparse Activation Compression for Ultra-long Sequence Modeling in Recommenders","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.IR","authors_text":"Binbin Cao, Changfa Wu, Daixuan Cheng, Jian Wu, Junyu Bi, Kun Yuan, Shuwen Xiao, Yuning Jiang","submitted_at":"2026-02-24T15:28:58Z","abstract_excerpt":"Modern recommender systems leverage ultra-long user behavior sequences to capture dynamic preferences, but end-to-end modeling is infeasible in production due to latency and memory constraints. While summarizing history via interest centers offers a practical alternative, existing methods struggle to (1) identify user-specific centers at appropriate granularity and (2) accurately assign behaviors, leading to quantization errors and loss of long-tail preferences. To alleviate these issues, we propose Hierarchical Sparse Activation Compression (HiSAC), an efficient framework for personalized seq"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.21009","kind":"arxiv","version":2},"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/2602.21009/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":"2602.21009","created_at":"2026-07-07T01:16:06.402310+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.21009v2","created_at":"2026-07-07T01:16:06.402310+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.21009","created_at":"2026-07-07T01:16:06.402310+00:00"},{"alias_kind":"pith_short_12","alias_value":"NHPKZ2KYHYYS","created_at":"2026-07-07T01:16:06.402310+00:00"},{"alias_kind":"pith_short_16","alias_value":"NHPKZ2KYHYYS63NW","created_at":"2026-07-07T01:16:06.402310+00:00"},{"alias_kind":"pith_short_8","alias_value":"NHPKZ2KY","created_at":"2026-07-07T01:16:06.402310+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.09888","citing_title":"SinkRec: Mitigating Semantic State Sink in Long Sequence Recommendation with Memory-Conditioned Gated Delta Networks","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2604.08933","citing_title":"IAT: Instance-As-Token Compression for Historical User Sequence Modeling in Industrial Recommender Systems","ref_index":44,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4","json":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4.json","graph_json":"https://pith.science/api/pith-number/NHPKZ2KYHYYS63NWRZE52XKFN4/graph.json","events_json":"https://pith.science/api/pith-number/NHPKZ2KYHYYS63NWRZE52XKFN4/events.json","paper":"https://pith.science/paper/NHPKZ2KY"},"agent_actions":{"view_html":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4","download_json":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4.json","view_paper":"https://pith.science/paper/NHPKZ2KY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.21009&json=true","fetch_graph":"https://pith.science/api/pith-number/NHPKZ2KYHYYS63NWRZE52XKFN4/graph.json","fetch_events":"https://pith.science/api/pith-number/NHPKZ2KYHYYS63NWRZE52XKFN4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4/action/storage_attestation","attest_author":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4/action/author_attestation","sign_citation":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4/action/citation_signature","submit_replication":"https://pith.science/pith/NHPKZ2KYHYYS63NWRZE52XKFN4/action/replication_record"}},"created_at":"2026-07-07T01:16:06.402310+00:00","updated_at":"2026-07-07T01:16:06.402310+00:00"}