{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:V7SWDBCX2C766G6PV24263LUQ3","short_pith_number":"pith:V7SWDBCX","schema_version":"1.0","canonical_sha256":"afe5618457d0bfef1bcfaeb9af6d7486e899a8d9f010a139dfe646120b903725","source":{"kind":"arxiv","id":"2505.11580","version":1},"attestation_state":"computed","paper":{"title":"Flash Invariant Point Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","q-bio.BM"],"primary_cat":"cs.LG","authors_text":"Andrew Liu, Axel Elaldi, Gurinder S Atwal, Nathan Russell, Nicholas T Franklin, Olivia Viessmann, Yih-En A Ban","submitted_at":"2025-05-16T16:19:05Z","abstract_excerpt":"Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-trainin"},"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":"2505.11580","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-16T16:19:05Z","cross_cats_sorted":["cs.AI","q-bio.BM"],"title_canon_sha256":"d5f88fe3ed65dd5b7982a7a27ecc7d822281fe2e65f3958d1cd138f971ccf395","abstract_canon_sha256":"d4eb12e142ae4dba84559e88ff4d1481f99e687276dd0019cdfbce6eb9009b20"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:04:23.958584Z","signature_b64":"yXL3eAs/Ad0I7Dg2xOmeUfYWrqrmURmn04FqfQFmvADvy+qWH2lQd7tN3BFylXQTtFzmWtZl8RqNGy6wB84mCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"afe5618457d0bfef1bcfaeb9af6d7486e899a8d9f010a139dfe646120b903725","last_reissued_at":"2026-07-05T11:04:23.958066Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:04:23.958066Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Flash Invariant Point Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","q-bio.BM"],"primary_cat":"cs.LG","authors_text":"Andrew Liu, Axel Elaldi, Gurinder S Atwal, Nathan Russell, Nicholas T Franklin, Olivia Viessmann, Yih-En A Ban","submitted_at":"2025-05-16T16:19:05Z","abstract_excerpt":"Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-trainin"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.11580","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/2505.11580/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":"2505.11580","created_at":"2026-07-05T11:04:23.958131+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.11580v1","created_at":"2026-07-05T11:04:23.958131+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.11580","created_at":"2026-07-05T11:04:23.958131+00:00"},{"alias_kind":"pith_short_12","alias_value":"V7SWDBCX2C76","created_at":"2026-07-05T11:04:23.958131+00:00"},{"alias_kind":"pith_short_16","alias_value":"V7SWDBCX2C766G6P","created_at":"2026-07-05T11:04:23.958131+00:00"},{"alias_kind":"pith_short_8","alias_value":"V7SWDBCX","created_at":"2026-07-05T11:04:23.958131+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.05439","citing_title":"Design-CP: Context Parallelism for Design of Protein Nanoparticles","ref_index":87,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3","json":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3.json","graph_json":"https://pith.science/api/pith-number/V7SWDBCX2C766G6PV24263LUQ3/graph.json","events_json":"https://pith.science/api/pith-number/V7SWDBCX2C766G6PV24263LUQ3/events.json","paper":"https://pith.science/paper/V7SWDBCX"},"agent_actions":{"view_html":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3","download_json":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3.json","view_paper":"https://pith.science/paper/V7SWDBCX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.11580&json=true","fetch_graph":"https://pith.science/api/pith-number/V7SWDBCX2C766G6PV24263LUQ3/graph.json","fetch_events":"https://pith.science/api/pith-number/V7SWDBCX2C766G6PV24263LUQ3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3/action/storage_attestation","attest_author":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3/action/author_attestation","sign_citation":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3/action/citation_signature","submit_replication":"https://pith.science/pith/V7SWDBCX2C766G6PV24263LUQ3/action/replication_record"}},"created_at":"2026-07-05T11:04:23.958131+00:00","updated_at":"2026-07-05T11:04:23.958131+00:00"}