{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WVB23FYORAZOABQYBGPFXNNSSW","short_pith_number":"pith:WVB23FYO","schema_version":"1.0","canonical_sha256":"b543ad970e8832e00618099e5bb5b29584ae3f120023aa09dcdf3f40754707b6","source":{"kind":"arxiv","id":"2503.03777","version":1},"attestation_state":"computed","paper":{"title":"FlexInfer: Breaking Memory Constraint via Flexible and Efficient Offloading for On-Device LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.OS","authors_text":"Arina Kharlamova, Chun Jason Xue, Hongchao Du, Nan Guan, Shangyu Wu","submitted_at":"2025-03-04T20:08:03Z","abstract_excerpt":"Large Language Models (LLMs) face challenges for on-device inference due to high memory demands. Traditional methods to reduce memory usage often compromise performance and lack adaptability. We propose FlexInfer, an optimized offloading framework for on-device inference, addressing these issues with techniques like asynchronous prefetching, balanced memory locking, and flexible tensor preservation. These strategies enhance memory efficiency and mitigate I/O bottlenecks, ensuring high performance within user-specified resource constraints. Experiments demonstrate that FlexInfer significantly i"},"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":"2503.03777","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.OS","submitted_at":"2025-03-04T20:08:03Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c793121406c141caba09be0e39d859569fd940dae76640dd4b4b2135726eb3b8","abstract_canon_sha256":"3b6156659dd838c13099565742c731ef53c29cca71edb5d9369cc609e7c51f97"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:25:20.938906Z","signature_b64":"heKD8cfjJh81m7QXLzX/3hsnRt2r4+V4/VzezMY3rJRnUjOUS6vuFF7RmJWyHUmFZJinKfGO3Ie8u9Za6T/ZDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b543ad970e8832e00618099e5bb5b29584ae3f120023aa09dcdf3f40754707b6","last_reissued_at":"2026-07-05T10:25:20.938334Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:25:20.938334Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FlexInfer: Breaking Memory Constraint via Flexible and Efficient Offloading for On-Device LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.OS","authors_text":"Arina Kharlamova, Chun Jason Xue, Hongchao Du, Nan Guan, Shangyu Wu","submitted_at":"2025-03-04T20:08:03Z","abstract_excerpt":"Large Language Models (LLMs) face challenges for on-device inference due to high memory demands. Traditional methods to reduce memory usage often compromise performance and lack adaptability. We propose FlexInfer, an optimized offloading framework for on-device inference, addressing these issues with techniques like asynchronous prefetching, balanced memory locking, and flexible tensor preservation. These strategies enhance memory efficiency and mitigate I/O bottlenecks, ensuring high performance within user-specified resource constraints. Experiments demonstrate that FlexInfer significantly i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.03777","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/2503.03777/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":"2503.03777","created_at":"2026-07-05T10:25:20.938409+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.03777v1","created_at":"2026-07-05T10:25:20.938409+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.03777","created_at":"2026-07-05T10:25:20.938409+00:00"},{"alias_kind":"pith_short_12","alias_value":"WVB23FYORAZO","created_at":"2026-07-05T10:25:20.938409+00:00"},{"alias_kind":"pith_short_16","alias_value":"WVB23FYORAZOABQY","created_at":"2026-07-05T10:25:20.938409+00:00"},{"alias_kind":"pith_short_8","alias_value":"WVB23FYO","created_at":"2026-07-05T10:25:20.938409+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27081","citing_title":"ReMoE: Boosting Expert Reuse through Router Fine-Tuning in Memory-Constrained MoE LLM Inference","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW","json":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW.json","graph_json":"https://pith.science/api/pith-number/WVB23FYORAZOABQYBGPFXNNSSW/graph.json","events_json":"https://pith.science/api/pith-number/WVB23FYORAZOABQYBGPFXNNSSW/events.json","paper":"https://pith.science/paper/WVB23FYO"},"agent_actions":{"view_html":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW","download_json":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW.json","view_paper":"https://pith.science/paper/WVB23FYO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.03777&json=true","fetch_graph":"https://pith.science/api/pith-number/WVB23FYORAZOABQYBGPFXNNSSW/graph.json","fetch_events":"https://pith.science/api/pith-number/WVB23FYORAZOABQYBGPFXNNSSW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW/action/storage_attestation","attest_author":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW/action/author_attestation","sign_citation":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW/action/citation_signature","submit_replication":"https://pith.science/pith/WVB23FYORAZOABQYBGPFXNNSSW/action/replication_record"}},"created_at":"2026-07-05T10:25:20.938409+00:00","updated_at":"2026-07-05T10:25:20.938409+00:00"}