{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:K7KSVAPYJZ3L7NWFA7DSDN2K7W","short_pith_number":"pith:K7KSVAPY","schema_version":"1.0","canonical_sha256":"57d52a81f84e76bfb6c507c721b74afdb7582946b2b713d7977bc9daa5e6d8bf","source":{"kind":"arxiv","id":"2607.18101","version":1},"attestation_state":"computed","paper":{"title":"Empowering On-Device Model Adaptation with an Edge AI Inference Accelerator","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AR","cs.CV"],"primary_cat":"cs.LG","authors_text":"Alessandro Capotondi, Marek Kraft, Mateusz Piechocki","submitted_at":"2026-07-20T16:02:58Z","abstract_excerpt":"On-device model adaptation is essential to enable lifelong personalization on resource-constrained hardware, but compute, power, and memory limitations of such devices make end-to-end backpropagation impractical for modern deep neural networks. This work proposes a heterogeneous adaptation pipeline that repurposes a commercial edge AI inference accelerator, Hailo-8L, for frozen-backbone feature extraction during on-device training. The computational graph is partitioned so that the pre-trained backbone is quantized to INT8 and run on the accelerator, while only a lightweight FP32 classificatio"},"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":"2607.18101","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-20T16:02:58Z","cross_cats_sorted":["cs.AR","cs.CV"],"title_canon_sha256":"cdcca81b4ae4686043effe5b3f15d5b625722712751f0fe2e1cdf3ea2eadf5e3","abstract_canon_sha256":"36b1192c1bac28c7e1ff5783a6e5153b59d7eaba4a939caeb1a318d9120c33a9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T02:22:16.419711Z","signature_b64":"4JpjtBjwy4bGzWcMb4TEHBjWaOarCHZ1Q0xpIRq0nA69fB49i9HFQRJL9BahWubphz/Tj8Xtfrg/3brYWVX4Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"57d52a81f84e76bfb6c507c721b74afdb7582946b2b713d7977bc9daa5e6d8bf","last_reissued_at":"2026-07-21T02:22:16.418893Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T02:22:16.418893Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Empowering On-Device Model Adaptation with an Edge AI Inference Accelerator","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AR","cs.CV"],"primary_cat":"cs.LG","authors_text":"Alessandro Capotondi, Marek Kraft, Mateusz Piechocki","submitted_at":"2026-07-20T16:02:58Z","abstract_excerpt":"On-device model adaptation is essential to enable lifelong personalization on resource-constrained hardware, but compute, power, and memory limitations of such devices make end-to-end backpropagation impractical for modern deep neural networks. This work proposes a heterogeneous adaptation pipeline that repurposes a commercial edge AI inference accelerator, Hailo-8L, for frozen-backbone feature extraction during on-device training. The computational graph is partitioned so that the pre-trained backbone is quantized to INT8 and run on the accelerator, while only a lightweight FP32 classificatio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.18101","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/2607.18101/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":"2607.18101","created_at":"2026-07-21T02:22:16.419309+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.18101v1","created_at":"2026-07-21T02:22:16.419309+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.18101","created_at":"2026-07-21T02:22:16.419309+00:00"},{"alias_kind":"pith_short_12","alias_value":"K7KSVAPYJZ3L","created_at":"2026-07-21T02:22:16.419309+00:00"},{"alias_kind":"pith_short_16","alias_value":"K7KSVAPYJZ3L7NWF","created_at":"2026-07-21T02:22:16.419309+00:00"},{"alias_kind":"pith_short_8","alias_value":"K7KSVAPY","created_at":"2026-07-21T02:22:16.419309+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/K7KSVAPYJZ3L7NWFA7DSDN2K7W","json":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W.json","graph_json":"https://pith.science/api/pith-number/K7KSVAPYJZ3L7NWFA7DSDN2K7W/graph.json","events_json":"https://pith.science/api/pith-number/K7KSVAPYJZ3L7NWFA7DSDN2K7W/events.json","paper":"https://pith.science/paper/K7KSVAPY"},"agent_actions":{"view_html":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W","download_json":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W.json","view_paper":"https://pith.science/paper/K7KSVAPY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.18101&json=true","fetch_graph":"https://pith.science/api/pith-number/K7KSVAPYJZ3L7NWFA7DSDN2K7W/graph.json","fetch_events":"https://pith.science/api/pith-number/K7KSVAPYJZ3L7NWFA7DSDN2K7W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W/action/storage_attestation","attest_author":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W/action/author_attestation","sign_citation":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W/action/citation_signature","submit_replication":"https://pith.science/pith/K7KSVAPYJZ3L7NWFA7DSDN2K7W/action/replication_record"}},"created_at":"2026-07-21T02:22:16.419309+00:00","updated_at":"2026-07-21T02:22:16.419309+00:00"}