{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:CCYRHP5GUMOWSPXXZN4ZQKQPTW","short_pith_number":"pith:CCYRHP5G","schema_version":"1.0","canonical_sha256":"10b113bfa6a31d693ef7cb79982a0f9d86b27afe8ab7b2aac828535a68a5f554","source":{"kind":"arxiv","id":"2310.17157","version":1},"attestation_state":"computed","paper":{"title":"Deja Vu: Contextual Sparsity for Efficient LLMs at Inference Time","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anshumali Shrivastava, Beidi Chen, Binhang Yuan, Ce Zhang, Christopher Re, Jue Wang, Tianyi Zhou, Tri Dao, Yuandong Tian, Zhao Song, Zichang Liu","submitted_at":"2023-10-26T05:01:09Z","abstract_excerpt":"Large language models (LLMs) with hundreds of billions of parameters have sparked a new wave of exciting AI applications. However, they are computationally expensive at inference time. Sparsity is a natural approach to reduce this cost, but existing methods either require costly retraining, have to forgo LLM's in-context learning ability, or do not yield wall-clock time speedup on modern hardware. We hypothesize that contextual sparsity, which are small, input-dependent sets of attention heads and MLP parameters that yield approximately the same output as the dense model for a given input, can"},"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":"2310.17157","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-10-26T05:01:09Z","cross_cats_sorted":[],"title_canon_sha256":"a2373d8298a6249c1540fe27bca747e5ad6d86b0090ca87607785be58ddcdac1","abstract_canon_sha256":"8f2c97e7e1e8daac2c8daec1e2dacdfeff9af44b31b93bac054bca6b39575a05"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:05:20.761187Z","signature_b64":"Tsm1ehF9q1E3BlHgfNFekmUmDOzfLJXS5S/fUf9Pr3n64O/prM3B4+rcz614L3lPNChliR78ptHUz886luHcCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"10b113bfa6a31d693ef7cb79982a0f9d86b27afe8ab7b2aac828535a68a5f554","last_reissued_at":"2026-07-05T07:05:20.760782Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:05:20.760782Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Deja Vu: Contextual Sparsity for Efficient LLMs at Inference Time","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Anshumali Shrivastava, Beidi Chen, Binhang Yuan, Ce Zhang, Christopher Re, Jue Wang, Tianyi Zhou, Tri Dao, Yuandong Tian, Zhao Song, Zichang Liu","submitted_at":"2023-10-26T05:01:09Z","abstract_excerpt":"Large language models (LLMs) with hundreds of billions of parameters have sparked a new wave of exciting AI applications. However, they are computationally expensive at inference time. Sparsity is a natural approach to reduce this cost, but existing methods either require costly retraining, have to forgo LLM's in-context learning ability, or do not yield wall-clock time speedup on modern hardware. We hypothesize that contextual sparsity, which are small, input-dependent sets of attention heads and MLP parameters that yield approximately the same output as the dense model for a given input, can"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.17157","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/2310.17157/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":"2310.17157","created_at":"2026-07-05T07:05:20.760837+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.17157v1","created_at":"2026-07-05T07:05:20.760837+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.17157","created_at":"2026-07-05T07:05:20.760837+00:00"},{"alias_kind":"pith_short_12","alias_value":"CCYRHP5GUMOW","created_at":"2026-07-05T07:05:20.760837+00:00"},{"alias_kind":"pith_short_16","alias_value":"CCYRHP5GUMOWSPXX","created_at":"2026-07-05T07:05:20.760837+00:00"},{"alias_kind":"pith_short_8","alias_value":"CCYRHP5G","created_at":"2026-07-05T07:05:20.760837+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07670","citing_title":"Does Bielik Know What It Doesn't Know? Activation Dispersion Separates Entity Familiarity from Factual Reliability Across Model Scale","ref_index":59,"is_internal_anchor":true},{"citing_arxiv_id":"2605.05899","citing_title":"VisMMOE: Exploiting Visual-Expert Affinity for Efficient Visual-Language MoE Offloading","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16883","citing_title":"SinkRouter: Sink-Aware Routing for Efficient Long-Context Decoding in Large Language and Multimodal Models","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW","json":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW.json","graph_json":"https://pith.science/api/pith-number/CCYRHP5GUMOWSPXXZN4ZQKQPTW/graph.json","events_json":"https://pith.science/api/pith-number/CCYRHP5GUMOWSPXXZN4ZQKQPTW/events.json","paper":"https://pith.science/paper/CCYRHP5G"},"agent_actions":{"view_html":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW","download_json":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW.json","view_paper":"https://pith.science/paper/CCYRHP5G","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.17157&json=true","fetch_graph":"https://pith.science/api/pith-number/CCYRHP5GUMOWSPXXZN4ZQKQPTW/graph.json","fetch_events":"https://pith.science/api/pith-number/CCYRHP5GUMOWSPXXZN4ZQKQPTW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW/action/storage_attestation","attest_author":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW/action/author_attestation","sign_citation":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW/action/citation_signature","submit_replication":"https://pith.science/pith/CCYRHP5GUMOWSPXXZN4ZQKQPTW/action/replication_record"}},"created_at":"2026-07-05T07:05:20.760837+00:00","updated_at":"2026-07-05T07:05:20.760837+00:00"}