{"paper":{"title":"Architecture Dependent Temporal Observability Under Deployment Interference in Edge Inference Systems","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Deployment interference can corrupt both inference timing and the software that measures it, independently.","cross_cats":["cs.SY"],"primary_cat":"eess.SY","authors_text":"Akul Swami, Nikhil Chougule","submitted_at":"2026-05-17T23:51:59Z","abstract_excerpt":"Edge inference systems are typically evaluated with software-reported latency collected under controlled conditions. We argue, and demonstrate empirically, that deployment interference can corrupt not only the inference timing being measured but the timing observability infrastructure that measures it, and that the two failures can occur independently.\n  We pair software-reported timing with externally observable GPIO intervals captured by a Saleae Logic Pro 8 logic analyzer on an NVIDIA Jetson Orin Nano, running MobileNetV2 under two inference architectures (TensorRT FP16 GPU and ONNX Runtime"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We claim, narrowly, that timing observability is itself an interference-sensitive resource, and that summary statistics from a single timing source can hide failure modes an independent external observer makes visible.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The experimental setup assumes that the Saleae Logic Pro 8 logic analyzer's GPIO interval captures provide a reliable, independent external ground truth that is not itself corrupted by the same deployment interference or pairing issues as the software reports.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Deployment interference corrupts timing observability in edge AI systems, allowing software logs to report normal operation while external hardware captures reveal failures that differ by inference architecture.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Deployment interference can corrupt both inference timing and the software that measures it, independently.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"b1d0f32356745cb468c6c05dd614fba66e1959fc7ef4001ceffb02a0dcad7b87"},"source":{"id":"2605.17701","kind":"arxiv","version":1},"verdict":{"id":"3fb7c6e7-4fa4-4e0e-bda5-defa3c4b1f4a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T22:01:43.557922Z","strongest_claim":"We claim, narrowly, that timing observability is itself an interference-sensitive resource, and that summary statistics from a single timing source can hide failure modes an independent external observer makes visible.","one_line_summary":"Deployment interference corrupts timing observability in edge AI systems, allowing software logs to report normal operation while external hardware captures reveal failures that differ by inference architecture.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The experimental setup assumes that the Saleae Logic Pro 8 logic analyzer's GPIO interval captures provide a reliable, independent external ground truth that is not itself corrupted by the same deployment interference or pairing issues as the software reports.","pith_extraction_headline":"Deployment interference can corrupt both inference timing and the software that measures it, independently."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.17701/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.408257Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:10:57.953630Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"shingle_duplication","ran_at":"2026-05-19T21:49:43.942023Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T21:49:43.740578Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.515019Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T21:21:59.480296Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.422998Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"62e4851ba6bf1b8f60b8ea4c469f246ac7a7109cff649fac9df8ca58be35be9d"},"references":{"count":11,"sample":[{"doi":"","year":null,"title":"and Reed, Daniel A","work_id":"02b1f630-d3cb-4dfe-8be2-d047d035d702","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Mytkowicz, Todd and Diwan, Amer and Hauswirth, Matthias and Sweeney, Peter F. , title =. Proceedings of the 14th International Conference on Architectural Support for Programming Languages and Operati","work_id":"e1749fac-5853-4e92-84ba-0955db34eb68","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"IEEE International Parallel and Distributed Processing Symposium , year =","work_id":"62da1229-21b5-4e3a-8b1a-a47b379753e9","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Ratul, I. J. and Zhou, Y. and Yang, K. , title =. Electronics , volume =","work_id":"aa6b87cb-11cb-4436-8e20-01414b373b60","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Don't Buy the Pig in a Poke: Benchmarking DNNs Inference Performance before Development , booktitle =","work_id":"1f4b1f45-705c-4a3b-bd64-ab17f6b8c02b","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":11,"snapshot_sha256":"b19761a8ac12f4903402e1935c5ebb8dd3997b25c3d6c1f0ee14ea297105150f","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"065d5ee850a6be6af68920d8c85f844670cef118a946e69f4ad75709bbcc7ac3"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}