{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:DVJ4QDECT33QSGJ4MCPMP4GBUT","short_pith_number":"pith:DVJ4QDEC","schema_version":"1.0","canonical_sha256":"1d53c80c829ef709193c609ec7f0c1a4c3abdb5c18d11fded41a650c2eed2d7a","source":{"kind":"arxiv","id":"2407.14057","version":1},"attestation_state":"computed","paper":{"title":"LazyLLM: Dynamic Token Pruning for Efficient Long Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Mahyar Najibi, Minsik Cho, Mohammad Rastegari, Qichen Fu, Sachin Mehta, Thomas Merth","submitted_at":"2024-07-19T06:34:45Z","abstract_excerpt":"The inference of transformer-based large language models consists of two sequential stages: 1) a prefilling stage to compute the KV cache of prompts and generate the first token, and 2) a decoding stage to generate subsequent tokens. For long prompts, the KV cache must be computed for all tokens during the prefilling stage, which can significantly increase the time needed to generate the first token. Consequently, the prefilling stage may become a bottleneck in the generation process. An open question remains whether all prompt tokens are essential for generating the first token. To answer thi"},"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":"2407.14057","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-19T06:34:45Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"6c25c748deced50f2d352d02e574cd6bf94df55f7a259ce0a873a818f694d0f2","abstract_canon_sha256":"b094743d553d1dc50c3d8b9cf46e3ab6a66eca5e346baef2ffa400e4ef7f53ba"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:45:57.828393Z","signature_b64":"XjAY4y3/DM1lar4eSXZ5m1uOyyLfb+ZCpdaaJozrUovzMcUe8rVu8tn5EPtXUNFxi5YYBa3II37hm3uOIn1NBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1d53c80c829ef709193c609ec7f0c1a4c3abdb5c18d11fded41a650c2eed2d7a","last_reissued_at":"2026-07-05T08:45:57.827844Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:45:57.827844Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LazyLLM: Dynamic Token Pruning for Efficient Long Context LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Mahyar Najibi, Minsik Cho, Mohammad Rastegari, Qichen Fu, Sachin Mehta, Thomas Merth","submitted_at":"2024-07-19T06:34:45Z","abstract_excerpt":"The inference of transformer-based large language models consists of two sequential stages: 1) a prefilling stage to compute the KV cache of prompts and generate the first token, and 2) a decoding stage to generate subsequent tokens. For long prompts, the KV cache must be computed for all tokens during the prefilling stage, which can significantly increase the time needed to generate the first token. Consequently, the prefilling stage may become a bottleneck in the generation process. An open question remains whether all prompt tokens are essential for generating the first token. To answer thi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.14057","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/2407.14057/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":"2407.14057","created_at":"2026-07-05T08:45:57.827905+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.14057v1","created_at":"2026-07-05T08:45:57.827905+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.14057","created_at":"2026-07-05T08:45:57.827905+00:00"},{"alias_kind":"pith_short_12","alias_value":"DVJ4QDECT33Q","created_at":"2026-07-05T08:45:57.827905+00:00"},{"alias_kind":"pith_short_16","alias_value":"DVJ4QDECT33QSGJ4","created_at":"2026-07-05T08:45:57.827905+00:00"},{"alias_kind":"pith_short_8","alias_value":"DVJ4QDEC","created_at":"2026-07-05T08:45:57.827905+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08057","citing_title":"Towards Efficient Large Language Model Serving: A Survey on System-Aware KV Cache Optimization","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2606.00523","citing_title":"ProactiveLLM: Learning Active Interaction for Streaming Large Language Models","ref_index":79,"is_internal_anchor":false},{"citing_arxiv_id":"2603.10726","citing_title":"PrefixWall: Mitigating Prefix Caching Side Channels in Shared LLM Systems","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2603.28458","citing_title":"HISA: Efficient Hierarchical Indexing for Fine-Grained Sparse Attention","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06554","citing_title":"Long Context Pre-Training with Lighthouse Attention","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06221","citing_title":"UniPrefill: Universal Long-Context Prefill Acceleration via Block-wise Dynamic Sparsification","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT","json":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT.json","graph_json":"https://pith.science/api/pith-number/DVJ4QDECT33QSGJ4MCPMP4GBUT/graph.json","events_json":"https://pith.science/api/pith-number/DVJ4QDECT33QSGJ4MCPMP4GBUT/events.json","paper":"https://pith.science/paper/DVJ4QDEC"},"agent_actions":{"view_html":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT","download_json":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT.json","view_paper":"https://pith.science/paper/DVJ4QDEC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.14057&json=true","fetch_graph":"https://pith.science/api/pith-number/DVJ4QDECT33QSGJ4MCPMP4GBUT/graph.json","fetch_events":"https://pith.science/api/pith-number/DVJ4QDECT33QSGJ4MCPMP4GBUT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT/action/storage_attestation","attest_author":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT/action/author_attestation","sign_citation":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT/action/citation_signature","submit_replication":"https://pith.science/pith/DVJ4QDECT33QSGJ4MCPMP4GBUT/action/replication_record"}},"created_at":"2026-07-05T08:45:57.827905+00:00","updated_at":"2026-07-05T08:45:57.827905+00:00"}