{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NFXWJCQPRPODMNIZPE7D5X44IN","short_pith_number":"pith:NFXWJCQP","schema_version":"1.0","canonical_sha256":"696f648a0f8bdc363519793e3edf9c434840adb6648ccfda9cb86eae6d1c12af","source":{"kind":"arxiv","id":"2608.08382","version":1},"attestation_state":"computed","paper":{"title":"LLMVisor: A Real-Time Latency Attribution Model for Multi-Tenant LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.AI","authors_text":"Jiaxin Shan, Le Xu, Liguang Xie, Shuowei Jin, Tieying Zhang, Xueshen Liu, Z. Morley Mao","submitted_at":"2026-08-09T00:28:05Z","abstract_excerpt":"As LLM inference shifts to multi-tenant GPU clusters, co-batching improves throughput but obscures per-tenant usage and limits control. Enabling fractional sharing of the inference engine requires a real-time, per-request attribution primitive that is accurate and light enough to run inside the scheduling loop. We present LLMVisor, a roofline-guided latency attribution model that captures the memory-bound and compute-bound phases via a concise piecewise-linear form over features proportional to FLOPs and memory I/O traffic. LLMVisor decomposes batch latency into additive, per-request shares an"},"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":"2608.08382","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-08-09T00:28:05Z","cross_cats_sorted":["cs.DC"],"title_canon_sha256":"5b1bde32f61c3cee015dc7e5faf8218f03b5c7405d1a75a30a0b73a532597d9f","abstract_canon_sha256":"734243d6d9936d560a4041648bc9f281d7513f71ee6e10b1dab9f30b05a4e0be"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-11T01:22:29.468681Z","signature_b64":"8ubhetiuM5XgSphaOHFLJuoeC15n+vjrCSjGfvjHJUi1fia8S783ZCyPGPC3dNNZyeKc4DoXspqIftqJWrz3Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"696f648a0f8bdc363519793e3edf9c434840adb6648ccfda9cb86eae6d1c12af","last_reissued_at":"2026-08-11T01:22:29.466400Z","signature_status":"signed_v1","first_computed_at":"2026-08-11T01:22:29.466400Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLMVisor: A Real-Time Latency Attribution Model for Multi-Tenant LLM Serving","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.DC"],"primary_cat":"cs.AI","authors_text":"Jiaxin Shan, Le Xu, Liguang Xie, Shuowei Jin, Tieying Zhang, Xueshen Liu, Z. Morley Mao","submitted_at":"2026-08-09T00:28:05Z","abstract_excerpt":"As LLM inference shifts to multi-tenant GPU clusters, co-batching improves throughput but obscures per-tenant usage and limits control. Enabling fractional sharing of the inference engine requires a real-time, per-request attribution primitive that is accurate and light enough to run inside the scheduling loop. We present LLMVisor, a roofline-guided latency attribution model that captures the memory-bound and compute-bound phases via a concise piecewise-linear form over features proportional to FLOPs and memory I/O traffic. LLMVisor decomposes batch latency into additive, per-request shares an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.08382","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/2608.08382/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":"2608.08382","created_at":"2026-08-11T01:22:29.467102+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.08382v1","created_at":"2026-08-11T01:22:29.467102+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.08382","created_at":"2026-08-11T01:22:29.467102+00:00"},{"alias_kind":"pith_short_12","alias_value":"NFXWJCQPRPOD","created_at":"2026-08-11T01:22:29.467102+00:00"},{"alias_kind":"pith_short_16","alias_value":"NFXWJCQPRPODMNIZ","created_at":"2026-08-11T01:22:29.467102+00:00"},{"alias_kind":"pith_short_8","alias_value":"NFXWJCQP","created_at":"2026-08-11T01:22:29.467102+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/NFXWJCQPRPODMNIZPE7D5X44IN","json":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN.json","graph_json":"https://pith.science/api/pith-number/NFXWJCQPRPODMNIZPE7D5X44IN/graph.json","events_json":"https://pith.science/api/pith-number/NFXWJCQPRPODMNIZPE7D5X44IN/events.json","paper":"https://pith.science/paper/NFXWJCQP"},"agent_actions":{"view_html":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN","download_json":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN.json","view_paper":"https://pith.science/paper/NFXWJCQP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.08382&json=true","fetch_graph":"https://pith.science/api/pith-number/NFXWJCQPRPODMNIZPE7D5X44IN/graph.json","fetch_events":"https://pith.science/api/pith-number/NFXWJCQPRPODMNIZPE7D5X44IN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN/action/storage_attestation","attest_author":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN/action/author_attestation","sign_citation":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN/action/citation_signature","submit_replication":"https://pith.science/pith/NFXWJCQPRPODMNIZPE7D5X44IN/action/replication_record"}},"created_at":"2026-08-11T01:22:29.467102+00:00","updated_at":"2026-08-11T01:22:29.467102+00:00"}