{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WVPUXISAULDBP7GD6YEILOSMGN","short_pith_number":"pith:WVPUXISA","schema_version":"1.0","canonical_sha256":"b55f4ba240a2c617fcc3f60885ba4c3345393579bcdcb24bbd2e0085cd2aa190","source":{"kind":"arxiv","id":"2605.15399","version":1},"attestation_state":"computed","paper":{"title":"Breakeven complexity: A new perspective on neural partial differential equation solvers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Neural PDE solvers reach cost parity with traditional methods after fewer solves as problems grow harder, higher-dimensional, or higher-Reynolds.","cross_cats":["cs.AI","cs.NA","math.NA","physics.comp-ph"],"primary_cat":"cs.LG","authors_text":"Mikhail Khodak, Nicholas Roberts, Tanya Marwah, Yijing Zhang","submitted_at":"2026-05-14T20:33:24Z","abstract_excerpt":"Neural surrogate solvers of partial differential equations (PDEs) promise dramatic speedups over numerical methods, especially in scenarios requiring many solves. However, current accuracy-based evaluations do not fully consider two central issues: (1) neural solvers incur substantial up-front costs for data generation, training, and tuning; and (2) classical solvers can also generate low-fidelity solutions at a sufficiently low simulation cost. To explicitly account for these realities and fully incorporate end-to-end costs, we propose an evaluation framework centered on breakeven complexity,"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.15399","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T20:33:24Z","cross_cats_sorted":["cs.AI","cs.NA","math.NA","physics.comp-ph"],"title_canon_sha256":"8e662b00ad6bc5c4121670df8f2ccac7cd1bc07754f22446887f74f502886c24","abstract_canon_sha256":"77471c699f1120652257fe9c4d64b281cb4922b846f4a061c7daa8ceda8bb13e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:00:56.665855Z","signature_b64":"85REIOQtUeZ3C+jXeABMY8ZDfh2oaQ1u0FZXU3E/NMIGYRFr6IkljGsz+DvQD8i+1L2iXDn5Hm4aMMex/tMqDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b55f4ba240a2c617fcc3f60885ba4c3345393579bcdcb24bbd2e0085cd2aa190","last_reissued_at":"2026-05-20T00:00:56.664930Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:00:56.664930Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Breakeven complexity: A new perspective on neural partial differential equation solvers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Neural PDE solvers reach cost parity with traditional methods after fewer solves as problems grow harder, higher-dimensional, or higher-Reynolds.","cross_cats":["cs.AI","cs.NA","math.NA","physics.comp-ph"],"primary_cat":"cs.LG","authors_text":"Mikhail Khodak, Nicholas Roberts, Tanya Marwah, Yijing Zhang","submitted_at":"2026-05-14T20:33:24Z","abstract_excerpt":"Neural surrogate solvers of partial differential equations (PDEs) promise dramatic speedups over numerical methods, especially in scenarios requiring many solves. However, current accuracy-based evaluations do not fully consider two central issues: (1) neural solvers incur substantial up-front costs for data generation, training, and tuning; and (2) classical solvers can also generate low-fidelity solutions at a sufficiently low simulation cost. To explicitly account for these realities and fully incorporate end-to-end costs, we propose an evaluation framework centered on breakeven complexity,"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"our results suggest that neural PDE solvers become more effective as problems get harder in terms of cost, dimension, rollout, physics regime (e.g. higher Reynolds number), etc.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That scaling laws can be reliably applied to allocate training budget between data generation and model training while achieving smooth error-matching between neural and traditional solvers across diverse PDE settings and regimes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Breakeven complexity is introduced to evaluate neural PDE solvers by total end-to-end cost, with results indicating they become advantageous for harder problems such as higher dimensions, longer rollouts, and higher Reynolds numbers.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Neural PDE solvers reach cost parity with traditional methods after fewer solves as problems grow harder, higher-dimensional, or higher-Reynolds.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f407785faaca8ed0f030bbc60478f4b0f0d97fd6450f5109f6b6b7cc406d22c1"},"source":{"id":"2605.15399","kind":"arxiv","version":1},"verdict":{"id":"fc1b075b-65ff-4ea0-b9b7-982483bf8d65","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T16:18:16.733337Z","strongest_claim":"our results suggest that neural PDE solvers become more effective as problems get harder in terms of cost, dimension, rollout, physics regime (e.g. higher Reynolds number), etc.","one_line_summary":"Breakeven complexity is introduced to evaluate neural PDE solvers by total end-to-end cost, with results indicating they become advantageous for harder problems such as higher dimensions, longer rollouts, and higher Reynolds numbers.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That scaling laws can be reliably applied to allocate training budget between data generation and model training while achieving smooth error-matching between neural and traditional solvers across diverse PDE settings and regimes.","pith_extraction_headline":"Neural PDE solvers reach cost parity with traditional methods after fewer solves as problems grow harder, higher-dimensional, or higher-Reynolds."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15399/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:31:18.271097Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T16:25:54.276643Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T16:23:54.965831Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T15:50:52.792080Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:21:54.161916Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.717417Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"342d1752755fc21b306c4d7130537ea95dbf09dfc9400a21349cb7e0e90238eb"},"references":{"count":38,"sample":[{"doi":"","year":2025,"title":"Abnar, S., H. Shah, D. Busbridge, A. El-Nouby, J. M. Susskind, and V . Thilak (2025). Parameters vs FLOPs: Scaling laws for optimal sparsity for mixture-of-experts language models. InForty- second Int","work_id":"c1654d31-3972-4a44-b0b8-20f037119662","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2005,"title":"Ahrens, J. P., B. Geveci, and C. C. Law (2005). ParaView: An end-user tool for large-data visualization. InThe Visualization Handbook","work_id":"a7c6e2c0-f674-4912-a2a7-a143d770c5aa","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Fluid intelligence: A forward look on AI foundation models in computational fluid dynamics","work_id":"4e0c155e-3ee4-4ac3-8847-0015a8b6e82c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2009,"title":"Bochev, P. B. and D. Ridzal (2009). An optimization-based approach for the design of pde solution algorithms.SIAM journal on numerical analysis 47(5), 3938–3955","work_id":"73ac52c6-0af9-4fa5-92f7-013d7338031b","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Choudhary, N., V . Singh, A. Talwalkar, N. M. Boffi, M. Khodak, and T. Marwah (2025). Pre-generating multi-difficulty pde data for few-shot neural pde solvers.arXiv preprint arXiv:2512.00564","work_id":"10a9b19a-8da3-475b-a6b6-b2450370ae1d","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":38,"snapshot_sha256":"7f2429e0267d059ca610d843408c10af7c45fdfc2909f10fe1551d130f2f5edd","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a24dbd72882ea802272649350d7343f71190d0073807c62e16bd412c523e27a6"},"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":"2605.15399","created_at":"2026-05-20T00:00:56.665089+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15399v1","created_at":"2026-05-20T00:00:56.665089+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15399","created_at":"2026-05-20T00:00:56.665089+00:00"},{"alias_kind":"pith_short_12","alias_value":"WVPUXISAULDB","created_at":"2026-05-20T00:00:56.665089+00:00"},{"alias_kind":"pith_short_16","alias_value":"WVPUXISAULDBP7GD","created_at":"2026-05-20T00:00:56.665089+00:00"},{"alias_kind":"pith_short_8","alias_value":"WVPUXISA","created_at":"2026-05-20T00:00:56.665089+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN","json":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN.json","graph_json":"https://pith.science/api/pith-number/WVPUXISAULDBP7GD6YEILOSMGN/graph.json","events_json":"https://pith.science/api/pith-number/WVPUXISAULDBP7GD6YEILOSMGN/events.json","paper":"https://pith.science/paper/WVPUXISA"},"agent_actions":{"view_html":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN","download_json":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN.json","view_paper":"https://pith.science/paper/WVPUXISA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15399&json=true","fetch_graph":"https://pith.science/api/pith-number/WVPUXISAULDBP7GD6YEILOSMGN/graph.json","fetch_events":"https://pith.science/api/pith-number/WVPUXISAULDBP7GD6YEILOSMGN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN/action/storage_attestation","attest_author":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN/action/author_attestation","sign_citation":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN/action/citation_signature","submit_replication":"https://pith.science/pith/WVPUXISAULDBP7GD6YEILOSMGN/action/replication_record"}},"created_at":"2026-05-20T00:00:56.665089+00:00","updated_at":"2026-05-20T00:00:56.665089+00:00"}