{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KITIJ35TOQZLX6IIDQP46OFKIL","short_pith_number":"pith:KITIJ35T","schema_version":"1.0","canonical_sha256":"522684efb37432bbf9081c1fcf38aa42f9fe0d16d7c3fc7e5925da7f38ef5c54","source":{"kind":"arxiv","id":"2506.14851","version":1},"attestation_state":"computed","paper":{"title":"Efficient Serving of LLM Applications with Probabilistic Demand Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Chen Chen, Minyi Guo, Shixuan Sun, Weiye Wang, Xusheng Chen, Yifei Liu, Yifei Zhu, Yizhou Shan, Zhenghao Gan, Zhenhua Han, Zuo Gan","submitted_at":"2025-06-17T03:49:25Z","abstract_excerpt":"Applications based on Large Language Models (LLMs) contains a series of tasks to address real-world problems with boosted capability, which have dynamic demand volumes on diverse backends. Existing serving systems treat the resource demands of LLM applications as a blackbox, compromising end-to-end efficiency due to improper queuing order and backend warm up latency. We find that the resource demands of LLM applications can be modeled in a general and accurate manner with Probabilistic Demand Graph (PDGraph). We then propose Hermes, which leverages PDGraph for efficient serving of LLM applicat"},"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":"2506.14851","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.DC","submitted_at":"2025-06-17T03:49:25Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"19837c73065ea53ad3a558f43811176d307fb203c467fef6e2e23e660d685846","abstract_canon_sha256":"da483c0a6d11bcbb1fe4f4b692d5dd47e60e3382cf6cdfc80ebd3bb43b28d099"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T02:16:52.605988Z","signature_b64":"UJaDe4eDX+aAdGrEcCOjrxv9H6LFpUJsWnZEk76jMHrHaG6c+gfZX0zAXsuJ06csVJRAkGZlRJI9TBSkExN0Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"522684efb37432bbf9081c1fcf38aa42f9fe0d16d7c3fc7e5925da7f38ef5c54","last_reissued_at":"2026-06-30T02:16:52.605368Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T02:16:52.605368Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Efficient Serving of LLM Applications with Probabilistic Demand Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.DC","authors_text":"Chen Chen, Minyi Guo, Shixuan Sun, Weiye Wang, Xusheng Chen, Yifei Liu, Yifei Zhu, Yizhou Shan, Zhenghao Gan, Zhenhua Han, Zuo Gan","submitted_at":"2025-06-17T03:49:25Z","abstract_excerpt":"Applications based on Large Language Models (LLMs) contains a series of tasks to address real-world problems with boosted capability, which have dynamic demand volumes on diverse backends. Existing serving systems treat the resource demands of LLM applications as a blackbox, compromising end-to-end efficiency due to improper queuing order and backend warm up latency. We find that the resource demands of LLM applications can be modeled in a general and accurate manner with Probabilistic Demand Graph (PDGraph). We then propose Hermes, which leverages PDGraph for efficient serving of LLM applicat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.14851","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/2506.14851/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":"2506.14851","created_at":"2026-06-30T02:16:52.605445+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.14851v1","created_at":"2026-06-30T02:16:52.605445+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.14851","created_at":"2026-06-30T02:16:52.605445+00:00"},{"alias_kind":"pith_short_12","alias_value":"KITIJ35TOQZL","created_at":"2026-06-30T02:16:52.605445+00:00"},{"alias_kind":"pith_short_16","alias_value":"KITIJ35TOQZLX6II","created_at":"2026-06-30T02:16:52.605445+00:00"},{"alias_kind":"pith_short_8","alias_value":"KITIJ35T","created_at":"2026-06-30T02:16:52.605445+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.12950","citing_title":"Maestro: Workload-Aware Cross-Cluster Scheduling for LLM-Based Multi-Agent Systems","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2510.18586","citing_title":"TokenCake: A KV-Cache-centric Serving Framework for LLM-based Multi-Agent Applications","ref_index":9,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL","json":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL.json","graph_json":"https://pith.science/api/pith-number/KITIJ35TOQZLX6IIDQP46OFKIL/graph.json","events_json":"https://pith.science/api/pith-number/KITIJ35TOQZLX6IIDQP46OFKIL/events.json","paper":"https://pith.science/paper/KITIJ35T"},"agent_actions":{"view_html":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL","download_json":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL.json","view_paper":"https://pith.science/paper/KITIJ35T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.14851&json=true","fetch_graph":"https://pith.science/api/pith-number/KITIJ35TOQZLX6IIDQP46OFKIL/graph.json","fetch_events":"https://pith.science/api/pith-number/KITIJ35TOQZLX6IIDQP46OFKIL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL/action/storage_attestation","attest_author":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL/action/author_attestation","sign_citation":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL/action/citation_signature","submit_replication":"https://pith.science/pith/KITIJ35TOQZLX6IIDQP46OFKIL/action/replication_record"}},"created_at":"2026-06-30T02:16:52.605445+00:00","updated_at":"2026-06-30T02:16:52.605445+00:00"}