{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7FT3R4XQLCTVWR523CFX6J7PK6","short_pith_number":"pith:7FT3R4XQ","schema_version":"1.0","canonical_sha256":"f967b8f2f058a75b47bad88b7f27ef579e869232b8c275fa670ebf8bd73fa033","source":{"kind":"arxiv","id":"2605.13343","version":2},"attestation_state":"computed","paper":{"title":"Hierarchical Transformer Preconditioning for Interactive Physics Simulation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A hierarchical transformer preconditioner solves stiff multiphase Poisson systems up to 28 times faster than standard GPU incomplete factorization.","cross_cats":["cs.DC","cs.LG","cs.NA","math.NA"],"primary_cat":"cs.GR","authors_text":"Carl Osborne, Crystal Owens, Minghao Guo, Wojciech Matusik","submitted_at":"2026-05-13T11:02:27Z","abstract_excerpt":"Neural preconditioners for real-time physics simulation offer promising data-driven priors, but they often fail to capture long-range couplings efficiently because they inherit local message passing or sparse-operator access patterns. We introduce the Hierarchical Transformer Preconditioner, a neural preconditioner anchored to a weak-admissibility H-matrix partition. The partition provides a multiscale structural prior (dense diagonal leaves plus coarsening off-diagonal tiles) that enables full-graph approximate-inverse computation with O(N) scaling at fixed block sizes. The network models the"},"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":"2605.13343","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.GR","submitted_at":"2026-05-13T11:02:27Z","cross_cats_sorted":["cs.DC","cs.LG","cs.NA","math.NA"],"title_canon_sha256":"b84f5f8f42baeca84db57e33ff3c38b3c6033eba735ed50b67230699d01c378f","abstract_canon_sha256":"7be0647020a3e2f740d0c52257ebe1424a07ad4192c39bca638bfe6a4f08e8eb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:48.367061Z","signature_b64":"TNJ/qaQgT2rcaJ15seElSEybLNYlsdW++DNp64325ssMM8AZff/tWnVk9ZIz9QCP2GsFmMxFj1BG3ar6nZGxDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f967b8f2f058a75b47bad88b7f27ef579e869232b8c275fa670ebf8bd73fa033","last_reissued_at":"2026-05-18T02:44:48.366686Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:48.366686Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hierarchical Transformer Preconditioning for Interactive Physics Simulation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"A hierarchical transformer preconditioner solves stiff multiphase Poisson systems up to 28 times faster than standard GPU incomplete factorization.","cross_cats":["cs.DC","cs.LG","cs.NA","math.NA"],"primary_cat":"cs.GR","authors_text":"Carl Osborne, Crystal Owens, Minghao Guo, Wojciech Matusik","submitted_at":"2026-05-13T11:02:27Z","abstract_excerpt":"Neural preconditioners for real-time physics simulation offer promising data-driven priors, but they often fail to capture long-range couplings efficiently because they inherit local message passing or sparse-operator access patterns. We introduce the Hierarchical Transformer Preconditioner, a neural preconditioner anchored to a weak-admissibility H-matrix partition. The partition provides a multiscale structural prior (dense diagonal leaves plus coarsening off-diagonal tiles) that enables full-graph approximate-inverse computation with O(N) scaling at fixed block sizes. The network models the"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On stiff multiphase Poisson systems (up to 100:1 density contrast, N = 1,024-16,384), the solver runs from ~143 to ~21 fps. At N = 8,192, it reaches 17.9 ms/frame, with 2.2x speedup over GPU Jacobi, ~28x over GPU IC/DILU (AMGX multicolor_dilu), and 2.7x over neural SPAI retrained per scale on the same benchmark.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The weak-admissibility H-matrix partition plus highway connections in the transformer are assumed to capture long-range couplings sufficiently for the cosine-Hutchinson objective to produce a preconditioner that improves PCG convergence on irregular spectra without post-hoc tuning.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"A hierarchical transformer preconditioner with H-matrix structure and cosine-Hutchinson training delivers up to 2.7x speedup over prior neural methods on stiff multiphase Poisson systems up to N=16384.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A hierarchical transformer preconditioner solves stiff multiphase Poisson systems up to 28 times faster than standard GPU incomplete factorization.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5d4cd04566f98bede90cf530938c2db68986170d94e5536a8f180fbf033cd4b5"},"source":{"id":"2605.13343","kind":"arxiv","version":2},"verdict":{"id":"11fc26fb-a73e-4ed9-8d93-e541e8b6a280","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T02:39:58.435457Z","strongest_claim":"On stiff multiphase Poisson systems (up to 100:1 density contrast, N = 1,024-16,384), the solver runs from ~143 to ~21 fps. At N = 8,192, it reaches 17.9 ms/frame, with 2.2x speedup over GPU Jacobi, ~28x over GPU IC/DILU (AMGX multicolor_dilu), and 2.7x over neural SPAI retrained per scale on the same benchmark.","one_line_summary":"A hierarchical transformer preconditioner with H-matrix structure and cosine-Hutchinson training delivers up to 2.7x speedup over prior neural methods on stiff multiphase Poisson systems up to N=16384.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The weak-admissibility H-matrix partition plus highway connections in the transformer are assumed to capture long-range couplings sufficiently for the cosine-Hutchinson objective to produce a preconditioner that improves PCG convergence on irregular spectra without post-hoc tuning.","pith_extraction_headline":"A hierarchical transformer preconditioner solves stiff multiphase Poisson systems up to 28 times faster than standard GPU incomplete factorization."},"references":{"count":3,"sample":[{"doi":"","year":2021,"title":"Swin Transformer: Hierarchical Vision Transformer using Shifted Windows","work_id":"d6d2cf36-1b48-40c9-86e8-1d5666324ae1","ref_index":1,"cited_arxiv_id":"2103.14030","is_internal_anchor":true},{"doi":"","year":2017,"title":"In Advances in Neural Information Processing Systems 30 (NIPS 2017)","work_id":"3f79b81f-78b7-4f3a-bb04-c8f238996bb8","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"https://arxiv.org/abs/2510.27517 5","work_id":"49a43e61-b266-4b12-9f83-41d8024d2a5b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":3,"snapshot_sha256":"f864819d64965bfa3de9eb0f1129a814a9d946b11a04f8ff5973a1e944973b57","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"2f05c689ad2a35c2b7371a74ecf6f193f0a15c9165ffdf79bd7dcc902a8f492b"},"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":"2605.13343","created_at":"2026-05-18T02:44:48.366742+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.13343v2","created_at":"2026-05-18T02:44:48.366742+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13343","created_at":"2026-05-18T02:44:48.366742+00:00"},{"alias_kind":"pith_short_12","alias_value":"7FT3R4XQLCTV","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"7FT3R4XQLCTVWR52","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"7FT3R4XQ","created_at":"2026-05-18T12:33:37.589309+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/7FT3R4XQLCTVWR523CFX6J7PK6","json":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6.json","graph_json":"https://pith.science/api/pith-number/7FT3R4XQLCTVWR523CFX6J7PK6/graph.json","events_json":"https://pith.science/api/pith-number/7FT3R4XQLCTVWR523CFX6J7PK6/events.json","paper":"https://pith.science/paper/7FT3R4XQ"},"agent_actions":{"view_html":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6","download_json":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6.json","view_paper":"https://pith.science/paper/7FT3R4XQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.13343&json=true","fetch_graph":"https://pith.science/api/pith-number/7FT3R4XQLCTVWR523CFX6J7PK6/graph.json","fetch_events":"https://pith.science/api/pith-number/7FT3R4XQLCTVWR523CFX6J7PK6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6/action/storage_attestation","attest_author":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6/action/author_attestation","sign_citation":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6/action/citation_signature","submit_replication":"https://pith.science/pith/7FT3R4XQLCTVWR523CFX6J7PK6/action/replication_record"}},"created_at":"2026-05-18T02:44:48.366742+00:00","updated_at":"2026-05-18T02:44:48.366742+00:00"}