{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:S4TM5VZDT7HOCFKKSVEHOMBZ74","short_pith_number":"pith:S4TM5VZD","schema_version":"1.0","canonical_sha256":"9726ced7239fcee1154a9548773039ff2ac8a882200f29a8bc8a3e88441de3b4","source":{"kind":"arxiv","id":"2503.20074","version":2},"attestation_state":"computed","paper":{"title":"Adaptive Orchestration for Large-Scale Inference on Heterogeneous Accelerator Systems Balancing Cost, Performance, and Resilience","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.PF","authors_text":"Imry Kissos, Yahav Biran","submitted_at":"2025-03-25T21:20:11Z","abstract_excerpt":"The surge in generative AI workloads has created a need for scalable inference systems that can flexibly harness both GPUs and specialized accelerators while containing operational costs. This paper proposes a hardware-agnostic control loop that adaptively allocates requests across heterogeneous accelerators based on real-time cost and capacity signals. The approach sustains low latency and high throughput by dynamically shifting between cost-optimized and capacity-optimized modes, ensuring the most efficient use of expensive compute resources under fluctuating availability. Evaluated using th"},"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":"2503.20074","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.PF","submitted_at":"2025-03-25T21:20:11Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"45d2ade8785927c3ed5bdced737ec5ba7286e380431bf2f4560e363f5e3a4d1b","abstract_canon_sha256":"99530a8cce1ca793128a274c2844c91c63c84e55ee8ce7e3f4cc3fd0b592ea79"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:40:18.941179Z","signature_b64":"lDQqcXGFiA3Sy93h2dfTyXKiCWJjSGF0wUux0HyIS2MUiqy1zFXRl2GoIuRmNIHWZIfpUlQ2Cxq/BdZgeUmQCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9726ced7239fcee1154a9548773039ff2ac8a882200f29a8bc8a3e88441de3b4","last_reissued_at":"2026-07-05T10:40:18.940709Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:40:18.940709Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adaptive Orchestration for Large-Scale Inference on Heterogeneous Accelerator Systems Balancing Cost, Performance, and Resilience","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.PF","authors_text":"Imry Kissos, Yahav Biran","submitted_at":"2025-03-25T21:20:11Z","abstract_excerpt":"The surge in generative AI workloads has created a need for scalable inference systems that can flexibly harness both GPUs and specialized accelerators while containing operational costs. This paper proposes a hardware-agnostic control loop that adaptively allocates requests across heterogeneous accelerators based on real-time cost and capacity signals. The approach sustains low latency and high throughput by dynamically shifting between cost-optimized and capacity-optimized modes, ensuring the most efficient use of expensive compute resources under fluctuating availability. Evaluated using th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.20074","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/2503.20074/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":"2503.20074","created_at":"2026-07-05T10:40:18.940763+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.20074v2","created_at":"2026-07-05T10:40:18.940763+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.20074","created_at":"2026-07-05T10:40:18.940763+00:00"},{"alias_kind":"pith_short_12","alias_value":"S4TM5VZDT7HO","created_at":"2026-07-05T10:40:18.940763+00:00"},{"alias_kind":"pith_short_16","alias_value":"S4TM5VZDT7HOCFKK","created_at":"2026-07-05T10:40:18.940763+00:00"},{"alias_kind":"pith_short_8","alias_value":"S4TM5VZD","created_at":"2026-07-05T10:40:18.940763+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.09194","citing_title":"Meta-Learning for Speeding Up Large Model Inference in Decentralized Environments","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74","json":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74.json","graph_json":"https://pith.science/api/pith-number/S4TM5VZDT7HOCFKKSVEHOMBZ74/graph.json","events_json":"https://pith.science/api/pith-number/S4TM5VZDT7HOCFKKSVEHOMBZ74/events.json","paper":"https://pith.science/paper/S4TM5VZD"},"agent_actions":{"view_html":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74","download_json":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74.json","view_paper":"https://pith.science/paper/S4TM5VZD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.20074&json=true","fetch_graph":"https://pith.science/api/pith-number/S4TM5VZDT7HOCFKKSVEHOMBZ74/graph.json","fetch_events":"https://pith.science/api/pith-number/S4TM5VZDT7HOCFKKSVEHOMBZ74/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74/action/timestamp_anchor","attest_storage":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74/action/storage_attestation","attest_author":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74/action/author_attestation","sign_citation":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74/action/citation_signature","submit_replication":"https://pith.science/pith/S4TM5VZDT7HOCFKKSVEHOMBZ74/action/replication_record"}},"created_at":"2026-07-05T10:40:18.940763+00:00","updated_at":"2026-07-05T10:40:18.940763+00:00"}