{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PH554IYEAGNRTM6TMNH77SAYJ6","short_pith_number":"pith:PH554IYE","schema_version":"1.0","canonical_sha256":"79fbde2304019b19b3d3634fffc8184f8eebd299fff86111ac6db5e77a624553","source":{"kind":"arxiv","id":"2602.13069","version":2},"attestation_state":"computed","paper":{"title":"Memory-Efficient Structured Backpropagation for On-Device LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jaeho Lee, Juneyoung Park, Seongwan Kim, Yuri Hong","submitted_at":"2026-02-13T16:24:33Z","abstract_excerpt":"On-device fine-tuning enables privacy-preserving personalization of large language models, but mobile devices impose severe memory constraints, typically 6--12GB shared across all workloads. Existing approaches force a trade-off between exact gradients with high memory (MeBP) and low memory with noisy estimates (MeZO). We propose Memory-efficient Structured Backpropagation (MeSP), which bridges this gap by manually deriving backward passes that exploit LoRA's low-rank structure. Our key insight is that the intermediate projection $h = xA$ can be recomputed during backward at minimal cost since"},"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":"2602.13069","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-13T16:24:33Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"0a6ab38e8a97bcdee4b1578ae48fab3deddc8961481f344dec6e373202151a23","abstract_canon_sha256":"cef5b4dc10b40afc15588b8651a8a4654f31062ebeb012c5d3e8006f460e9907"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-01T01:02:34.314846Z","signature_b64":"6usYCilTKmbkDf1cK3UtlXurlMGr5F79dS+Ovq+XpOCo1Y3oxyHtLKbXxdLAT315UOUi+pUvRY4bvdnYleAaAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"79fbde2304019b19b3d3634fffc8184f8eebd299fff86111ac6db5e77a624553","last_reissued_at":"2026-06-01T01:02:34.313957Z","signature_status":"signed_v1","first_computed_at":"2026-06-01T01:02:34.313957Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Memory-Efficient Structured Backpropagation for On-Device LLM Fine-Tuning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jaeho Lee, Juneyoung Park, Seongwan Kim, Yuri Hong","submitted_at":"2026-02-13T16:24:33Z","abstract_excerpt":"On-device fine-tuning enables privacy-preserving personalization of large language models, but mobile devices impose severe memory constraints, typically 6--12GB shared across all workloads. Existing approaches force a trade-off between exact gradients with high memory (MeBP) and low memory with noisy estimates (MeZO). We propose Memory-efficient Structured Backpropagation (MeSP), which bridges this gap by manually deriving backward passes that exploit LoRA's low-rank structure. Our key insight is that the intermediate projection $h = xA$ can be recomputed during backward at minimal cost since"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.13069","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/2602.13069/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":"2602.13069","created_at":"2026-06-01T01:02:34.314087+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.13069v2","created_at":"2026-06-01T01:02:34.314087+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.13069","created_at":"2026-06-01T01:02:34.314087+00:00"},{"alias_kind":"pith_short_12","alias_value":"PH554IYEAGNR","created_at":"2026-06-01T01:02:34.314087+00:00"},{"alias_kind":"pith_short_16","alias_value":"PH554IYEAGNRTM6T","created_at":"2026-06-01T01:02:34.314087+00:00"},{"alias_kind":"pith_short_8","alias_value":"PH554IYE","created_at":"2026-06-01T01:02:34.314087+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6","json":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6.json","graph_json":"https://pith.science/api/pith-number/PH554IYEAGNRTM6TMNH77SAYJ6/graph.json","events_json":"https://pith.science/api/pith-number/PH554IYEAGNRTM6TMNH77SAYJ6/events.json","paper":"https://pith.science/paper/PH554IYE"},"agent_actions":{"view_html":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6","download_json":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6.json","view_paper":"https://pith.science/paper/PH554IYE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.13069&json=true","fetch_graph":"https://pith.science/api/pith-number/PH554IYEAGNRTM6TMNH77SAYJ6/graph.json","fetch_events":"https://pith.science/api/pith-number/PH554IYEAGNRTM6TMNH77SAYJ6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6/action/storage_attestation","attest_author":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6/action/author_attestation","sign_citation":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6/action/citation_signature","submit_replication":"https://pith.science/pith/PH554IYEAGNRTM6TMNH77SAYJ6/action/replication_record"}},"created_at":"2026-06-01T01:02:34.314087+00:00","updated_at":"2026-06-01T01:02:34.314087+00:00"}