{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PODFNQCBJSKNXODMSWDX2RMQTD","short_pith_number":"pith:PODFNQCB","schema_version":"1.0","canonical_sha256":"7b8656c0414c94dbb86c95877d459098f7aa6d8e18d835fc609c2f9388500b3b","source":{"kind":"arxiv","id":"2604.00499","version":2},"attestation_state":"computed","paper":{"title":"Scheduling LLM Inference with Uncertainty-Aware Output Length Predictions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Fangcheng Fu, Hao Luo, Hao Wang, Haoyu Zheng, Hongchao Zhu, Jiawei Jiang, Xiaokai Zhou, Xiao Yan, Yongqiang Zhang, Yuanyuan Zhu","submitted_at":"2026-04-01T05:31:21Z","abstract_excerpt":"To schedule LLM inference, the \\textit{shortest job first} (SJF) principle is favorable by prioritizing requests with short output lengths to avoid head-of-line (HOL) blocking. Existing methods usually predict a single output length for each request to facilitate scheduling. We argue that such a \\textit{point estimate} does not match the \\textit{stochastic} decoding process of LLM inference, where output length is \\textit{uncertain} by nature and determined by when the end-of-sequence (EOS) token is sampled. Hence, the output length of each request should be fitted with a distribution rather t"},"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":"2604.00499","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-04-01T05:31:21Z","cross_cats_sorted":[],"title_canon_sha256":"e994b8631553cdf31074fe20d7851ca6127853a8dfef1e1c34a0e0eaaebf7d20","abstract_canon_sha256":"82db838fb7503e473c77aa22badf13cc6e6e5eb628b6c06164529fd3f8d885f3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:05:08.721891Z","signature_b64":"B8lC+PdzufxhCcg3mNbxZC63lOr6UHMWQHqiwtuwFkZmfDkd1yzpP7pY/Cx9GP4s9w4RpftVchbzz+G0U94yDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7b8656c0414c94dbb86c95877d459098f7aa6d8e18d835fc609c2f9388500b3b","last_reissued_at":"2026-05-26T02:05:08.720791Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:05:08.720791Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scheduling LLM Inference with Uncertainty-Aware Output Length Predictions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Fangcheng Fu, Hao Luo, Hao Wang, Haoyu Zheng, Hongchao Zhu, Jiawei Jiang, Xiaokai Zhou, Xiao Yan, Yongqiang Zhang, Yuanyuan Zhu","submitted_at":"2026-04-01T05:31:21Z","abstract_excerpt":"To schedule LLM inference, the \\textit{shortest job first} (SJF) principle is favorable by prioritizing requests with short output lengths to avoid head-of-line (HOL) blocking. Existing methods usually predict a single output length for each request to facilitate scheduling. We argue that such a \\textit{point estimate} does not match the \\textit{stochastic} decoding process of LLM inference, where output length is \\textit{uncertain} by nature and determined by when the end-of-sequence (EOS) token is sampled. Hence, the output length of each request should be fitted with a distribution rather t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.00499","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/2604.00499/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":"2604.00499","created_at":"2026-05-26T02:05:08.720942+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.00499v2","created_at":"2026-05-26T02:05:08.720942+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.00499","created_at":"2026-05-26T02:05:08.720942+00:00"},{"alias_kind":"pith_short_12","alias_value":"PODFNQCBJSKN","created_at":"2026-05-26T02:05:08.720942+00:00"},{"alias_kind":"pith_short_16","alias_value":"PODFNQCBJSKNXODM","created_at":"2026-05-26T02:05:08.720942+00:00"},{"alias_kind":"pith_short_8","alias_value":"PODFNQCB","created_at":"2026-05-26T02:05:08.720942+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.07931","citing_title":"Robust Length Prediction: A Perspective from Heavy-Tailed Prompt-Conditioned Distributions","ref_index":21,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD","json":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD.json","graph_json":"https://pith.science/api/pith-number/PODFNQCBJSKNXODMSWDX2RMQTD/graph.json","events_json":"https://pith.science/api/pith-number/PODFNQCBJSKNXODMSWDX2RMQTD/events.json","paper":"https://pith.science/paper/PODFNQCB"},"agent_actions":{"view_html":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD","download_json":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD.json","view_paper":"https://pith.science/paper/PODFNQCB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.00499&json=true","fetch_graph":"https://pith.science/api/pith-number/PODFNQCBJSKNXODMSWDX2RMQTD/graph.json","fetch_events":"https://pith.science/api/pith-number/PODFNQCBJSKNXODMSWDX2RMQTD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD/action/storage_attestation","attest_author":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD/action/author_attestation","sign_citation":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD/action/citation_signature","submit_replication":"https://pith.science/pith/PODFNQCBJSKNXODMSWDX2RMQTD/action/replication_record"}},"created_at":"2026-05-26T02:05:08.720942+00:00","updated_at":"2026-05-26T02:05:08.720942+00:00"}