{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QVNYVYOTTO26MHICNV2AP5OZF6","short_pith_number":"pith:QVNYVYOT","schema_version":"1.0","canonical_sha256":"855b8ae1d39bb5e61d026d7407f5d92f8bb8b25993ecd93b29a883e1b452c52f","source":{"kind":"arxiv","id":"2505.21594","version":1},"attestation_state":"computed","paper":{"title":"Fast and Cost-effective Speculative Edge-Cloud Decoding with Early Exits","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC"],"primary_cat":"cs.RO","authors_text":"Priyadarshini Panda, Souvik Kundu, Yeshwanth Venkatesha","submitted_at":"2025-05-27T14:55:16Z","abstract_excerpt":"Large Language Models (LLMs) enable various applications on edge devices such as smartphones, wearables, and embodied robots. However, their deployment often depends on expensive cloud-based APIs, creating high operational costs, which limit access for smaller organizations and raise sustainability concerns. Certain LLMs can be deployed on-device, offering a cost-effective solution with reduced latency and improved privacy. Yet, limited computing resources constrain the size and accuracy of models that can be deployed, necessitating a collaborative design between edge and cloud. We propose a f"},"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":"2505.21594","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.RO","submitted_at":"2025-05-27T14:55:16Z","cross_cats_sorted":["cs.AI","cs.DC"],"title_canon_sha256":"f0f0d3782a773439645881fca55292958349ba384d4c7e6372ea4c6ab16c9262","abstract_canon_sha256":"7754a1b10369079668b1ebb37672b67935db00ac65aafaa7a189549a4036c677"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:00.651461Z","signature_b64":"pnVCjj9keXWPwO9yNwvYL9lPkCZOodqhaBSLNm1Mgs5MalzhVXzZ6QryasbyyU9zBFGbypWhLdk8ZR0UuA11AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"855b8ae1d39bb5e61d026d7407f5d92f8bb8b25993ecd93b29a883e1b452c52f","last_reissued_at":"2026-07-05T11:11:00.650972Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:00.650972Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fast and Cost-effective Speculative Edge-Cloud Decoding with Early Exits","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.DC"],"primary_cat":"cs.RO","authors_text":"Priyadarshini Panda, Souvik Kundu, Yeshwanth Venkatesha","submitted_at":"2025-05-27T14:55:16Z","abstract_excerpt":"Large Language Models (LLMs) enable various applications on edge devices such as smartphones, wearables, and embodied robots. However, their deployment often depends on expensive cloud-based APIs, creating high operational costs, which limit access for smaller organizations and raise sustainability concerns. Certain LLMs can be deployed on-device, offering a cost-effective solution with reduced latency and improved privacy. Yet, limited computing resources constrain the size and accuracy of models that can be deployed, necessitating a collaborative design between edge and cloud. We propose a f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.21594","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/2505.21594/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":"2505.21594","created_at":"2026-07-05T11:11:00.651035+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.21594v1","created_at":"2026-07-05T11:11:00.651035+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.21594","created_at":"2026-07-05T11:11:00.651035+00:00"},{"alias_kind":"pith_short_12","alias_value":"QVNYVYOTTO26","created_at":"2026-07-05T11:11:00.651035+00:00"},{"alias_kind":"pith_short_16","alias_value":"QVNYVYOTTO26MHIC","created_at":"2026-07-05T11:11:00.651035+00:00"},{"alias_kind":"pith_short_8","alias_value":"QVNYVYOT","created_at":"2026-07-05T11:11:00.651035+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2601.11652","citing_title":"WISP: Waste- and Interference-Suppressed Distributed Speculative LLM Serving at the Edge via Dynamic Drafting and SLO-Aware Batching","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10124","citing_title":"GELATO: Generative Entropy- and Lyapunov-based Adaptive Token Offloading for Device-Edge Speculative LLM Inference","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20919","citing_title":"DiP-SD: Distributed Pipelined Speculative Decoding for Efficient LLM Inference at the Edge","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07857","citing_title":"Networking-Aware Energy Efficiency in Agentic AI Inference: A Survey","ref_index":108,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6","json":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6.json","graph_json":"https://pith.science/api/pith-number/QVNYVYOTTO26MHICNV2AP5OZF6/graph.json","events_json":"https://pith.science/api/pith-number/QVNYVYOTTO26MHICNV2AP5OZF6/events.json","paper":"https://pith.science/paper/QVNYVYOT"},"agent_actions":{"view_html":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6","download_json":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6.json","view_paper":"https://pith.science/paper/QVNYVYOT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.21594&json=true","fetch_graph":"https://pith.science/api/pith-number/QVNYVYOTTO26MHICNV2AP5OZF6/graph.json","fetch_events":"https://pith.science/api/pith-number/QVNYVYOTTO26MHICNV2AP5OZF6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6/action/storage_attestation","attest_author":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6/action/author_attestation","sign_citation":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6/action/citation_signature","submit_replication":"https://pith.science/pith/QVNYVYOTTO26MHICNV2AP5OZF6/action/replication_record"}},"created_at":"2026-07-05T11:11:00.651035+00:00","updated_at":"2026-07-05T11:11:00.651035+00:00"}