{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:MTV2REHBQCXBXFOAV5WD52WAMC","merge_version":"pith-open-graph-merge-v1","event_count":4,"valid_event_count":4,"invalid_event_count":0,"equivocation_count":1,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"a93c34770cedaecb9472883ae67572009cbfa07af587427f27afef18ebf37f0a","cross_cats_sorted":["physics.comp-ph"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CE","submitted_at":"2026-05-13T11:34:32Z","title_canon_sha256":"708eb2ce55b461502af8ba7190e9fdffa4398860a895dbe0334c4038d4e7dcde"},"schema_version":"1.0","source":{"id":"2605.13378","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13378","created_at":"2026-05-18T02:44:47Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13378v1","created_at":"2026-05-18T02:44:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13378","created_at":"2026-05-18T02:44:47Z"},{"alias_kind":"pith_short_12","alias_value":"MTV2REHBQCXB","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"MTV2REHBQCXBXFOA","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"MTV2REHB","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:9d64534d2baf6c563ec66c5e9056f99e6a9b2e51750eb5c84672f87f378c2f4b","target":"graph","created_at":"2026-05-18T02:44:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"By preventing degradation of the Krylov operator, AD accelerates computation by 2-3 orders of magnitude across both CPU and GPU architectures. More importantly, it drastically improves global solver robustness, achieving a minimum completion rate of 95%, compared to just 42% for FD."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the observed gains are caused solely by the accuracy of the Gateaux derivatives and will persist when the automatic-differentiation implementation, problem size, or floating-point precision changes."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Forward-mode automatic differentiation replaces finite-difference approximations for Jacobian-vector products in JFNK solvers, delivering 2-3 orders of magnitude speedup and lifting minimum solver completion from 42% to 95% across Burgers, radiation diffusion, reaction-diffusion, and nonlinear time-"},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Automatic differentiation for Jacobian-vector products makes matrix-free Newton-Krylov solvers orders of magnitude faster and far more robust than finite differences."}],"snapshot_sha256":"98da9c81a3478bc72be14c42e5543a0675e7ac4c384575e24fab7f4657d00f07"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2e693b03efbeb288ee1e1d14ce2123f1d1a1f0374f563cb582f9aca697380c30"},"paper":{"abstract_excerpt":"Jacobian-Free Newton-Krylov (JFNK) methods avoid forming the full Jacobian, but still require Jacobian-vector products, i.e., Gateaux derivatives of the nonlinear residual along Krylov directions. In standard Finite Differences (FD) formulations, these products are obtained by perturbing the Newton state and differencing residuals, making the linearization sensitive to round-off error and floating-point precision. This work evaluates the global impact of forward-mode Automatic Differentiation (AD) as a replacement for FD Jacobian-vector product in finite-precision JFNK solvers. The comparison ","authors_text":"Marco Pasquale, Stefano Markidis","cross_cats":["physics.comp-ph"],"headline":"Automatic differentiation for Jacobian-vector products makes matrix-free Newton-Krylov solvers orders of magnitude faster and far more robust than finite differences.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CE","submitted_at":"2026-05-13T11:34:32Z","title":"Robust Matrix-Free Newton-Krylov Solvers via Automatic Differentiation"},"references":{"count":31,"internal_anchors":1,"resolved_work":31,"sample":[{"cited_arxiv_id":"","doi":"10.1016/j.jcp.2003.08.010","is_internal_anchor":false,"ref_index":1,"title":"D. Knoll, D. Keyes, Jacobian-free Newton–Krylov methods: a survey of approaches and applications, Journal of Computational Physics 193 (2) 26 (2004) 357–397.doi:10.1016/j.jcp.2003.08.010. URLhttps://l","work_id":"0af1b009-f84f-4a2c-a4e6-bc1cfd9f18d9","year":2004},{"cited_arxiv_id":"","doi":"10.1137/1.9781611970944","is_internal_anchor":false,"ref_index":2,"title":"C. T. Kelley, Iterative Methods for Linear and Nonlinear Equations, Frontiers in Applied Mathematics, Society for Industrial and Applied Mathematics, 1995.doi:10.1137/1.9781611970944. URLhttps://epubs","work_id":"9fc8315a-16e1-43fe-a09f-dd166e2f7178","year":1995},{"cited_arxiv_id":"","doi":"10.1137/0907058","is_internal_anchor":false,"ref_index":3,"title":"GMRES: A generalized minimal residual algorithm for solving nonsymmetric linear systems","work_id":"f6da0283-d7be-49a1-9588-ef5e70259b34","year":1986},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"M. R. Hestenes, E. Stiefel, Methods of Conjugate Gradients for Solving LinearSystems, JournalofResearchoftheNationalBureauofStandards 49 (6) (1952) 409–436","work_id":"657d203e-cd14-4080-8004-d235c4806e38","year":1952},{"cited_arxiv_id":"","doi":"10.1137/0913035","is_internal_anchor":false,"ref_index":5,"title":"H. A. Van Der Vorst, Bi-CGSTAB: A Fast and Smoothly Converging Variant of Bi-CG for the Solution of Nonsymmetric Linear Systems, SIAM Journal on Scientific and Statistical Computing 13 (2) (1992) 631–","work_id":"77df7389-dd91-4e7f-b294-eb5f25f03be0","year":1992}],"snapshot_sha256":"3c3ab9e0dc756420ff341358b00bbe2e135fcb6a18bd800ececd7cfc07a25d6f"},"source":{"id":"2605.13378","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:53:47.699379Z","id":"84ca345b-833d-418c-acce-13e3866dd439","model_set":{"reader":"grok-4.3"},"one_line_summary":"Forward-mode automatic differentiation replaces finite-difference approximations for Jacobian-vector products in JFNK solvers, delivering 2-3 orders of magnitude speedup and lifting minimum solver completion from 42% to 95% across Burgers, radiation diffusion, reaction-diffusion, and nonlinear time-","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Automatic differentiation for Jacobian-vector products makes matrix-free Newton-Krylov solvers orders of magnitude faster and far more robust than finite differences.","strongest_claim":"By preventing degradation of the Krylov operator, AD accelerates computation by 2-3 orders of magnitude across both CPU and GPU architectures. More importantly, it drastically improves global solver robustness, achieving a minimum completion rate of 95%, compared to just 42% for FD.","weakest_assumption":"That the observed gains are caused solely by the accuracy of the Gateaux derivatives and will persist when the automatic-differentiation implementation, problem size, or floating-point precision changes."}},"verdict_id":"84ca345b-833d-418c-acce-13e3866dd439"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:f550cecc73cb4570d21c35bcd79118b2d5ca204cff4ee1ab958aa0b2fd96ba80","target":"record","created_at":"2026-05-18T02:44:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"a93c34770cedaecb9472883ae67572009cbfa07af587427f27afef18ebf37f0a","cross_cats_sorted":["physics.comp-ph"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CE","submitted_at":"2026-05-13T11:34:32Z","title_canon_sha256":"708eb2ce55b461502af8ba7190e9fdffa4398860a895dbe0334c4038d4e7dcde"},"schema_version":"1.0","source":{"id":"2605.13378","kind":"arxiv","version":1}},"canonical_sha256":"64eba890e180ae1b95c0af6c3eeac060b2c692adf076a2af7267d51c48ee6fd8","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"64eba890e180ae1b95c0af6c3eeac060b2c692adf076a2af7267d51c48ee6fd8","first_computed_at":"2026-05-18T02:44:47.873733Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T02:44:47.873733Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"0ipRyw1t0YcgnbbrZxUL/sfDfiUD86kIfXK3BtqJG1DUo8QIiW0rXMujHULhfNYtzDm3fRKdytUTkp2Vf60DBQ==","signature_status":"signed_v1","signed_at":"2026-05-18T02:44:47.874253Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13378","source_kind":"arxiv","source_version":1}}},"equivocations":[{"signer_id":"pith.science","event_type":"integrity_finding","target":"integrity","event_ids":["sha256:78e6dae54b8bb6f341a13cb6ddc1fa42833c5526beabaf389fdd87a548710472","sha256:ff7ae82fac0ae61850aa49a7bf35359961bed729afc4d81a53634e1590aedb94"]}],"invalid_events":[],"applied_event_ids":["sha256:f550cecc73cb4570d21c35bcd79118b2d5ca204cff4ee1ab958aa0b2fd96ba80","sha256:9d64534d2baf6c563ec66c5e9056f99e6a9b2e51750eb5c84672f87f378c2f4b"],"state_sha256":"85980de23be1ffa714b4cdcaee12ef8806058eb46d82b643141b8b2300a7f0fb"}