{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:SRQXSWF6VX2ACCUS64ZF2DZQBP","short_pith_number":"pith:SRQXSWF6","canonical_record":{"source":{"id":"2603.02337","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-02T19:09:15Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"e4433be4b722edab0843c62d68496fafab0d6be6ca150dacc1a83daa51daba46","abstract_canon_sha256":"f82d9cbde890954c96bb660ae47ae3beffc3e8acf2003284ad1a1f3f4f704c72"},"schema_version":"1.0"},"canonical_sha256":"94617958beadf4010a92f7325d0f300be322ebd7e4bd33b1db94139de227c6e5","source":{"kind":"arxiv","id":"2603.02337","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.02337","created_at":"2026-05-18T03:09:23Z"},{"alias_kind":"arxiv_version","alias_value":"2603.02337v2","created_at":"2026-05-18T03:09:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.02337","created_at":"2026-05-18T03:09:23Z"},{"alias_kind":"pith_short_12","alias_value":"SRQXSWF6VX2A","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"SRQXSWF6VX2ACCUS","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"SRQXSWF6","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:SRQXSWF6VX2ACCUS64ZF2DZQBP","target":"record","payload":{"canonical_record":{"source":{"id":"2603.02337","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-02T19:09:15Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"e4433be4b722edab0843c62d68496fafab0d6be6ca150dacc1a83daa51daba46","abstract_canon_sha256":"f82d9cbde890954c96bb660ae47ae3beffc3e8acf2003284ad1a1f3f4f704c72"},"schema_version":"1.0"},"canonical_sha256":"94617958beadf4010a92f7325d0f300be322ebd7e4bd33b1db94139de227c6e5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:09:23.167582Z","signature_b64":"N7jTCLWitaAYXplYzZAZ8ScEl3e/tMPATMfAqVlnBoMLcH7/nPO2N9XG0CooH5AuG4q4qf4Cghm34SwyDlraBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"94617958beadf4010a92f7325d0f300be322ebd7e4bd33b1db94139de227c6e5","last_reissued_at":"2026-05-18T03:09:23.166936Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:09:23.166936Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2603.02337","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:09:23Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"T9qS9QuLNOfkJvVn95/lXu1LEkE5ibYBBs6Yw9arCk/Pwko1C8lDXA/sSueWZQLO8n1nXMMW541amIO49AG6CQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-22T23:31:05.229142Z"},"content_sha256":"5bd665e45ea4c59699646aa546de887bbb84aee6fdcbf2dd78fb66fccf7e8ea3","schema_version":"1.0","event_id":"sha256:5bd665e45ea4c59699646aa546de887bbb84aee6fdcbf2dd78fb66fccf7e8ea3"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:SRQXSWF6VX2ACCUS64ZF2DZQBP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Preconditioned Flow Matching","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Preconditioning transforms targets to isotropic space to reshape flow matching paths and fix ill-conditioned optimization.","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Eldad Haber, Eshed Gal, Md Shahriar Rahim Siddiqui, Moshe Eliasof, Shadab Ahamed, Simon Ghyselincks","submitted_at":"2026-03-02T19:09:15Z","abstract_excerpt":"Flow matching (FM) learns vector fields by regressing stochastic velocity targets along intermediate distributions $p_t$. We identify a geometric optimization bottleneck in this regression problem: when the covariance $\\Sigma_t$ of $p_t$ is ill-conditioned, gradient-based training rapidly fits high-variance directions while making slow progress along low-variance ones. In an exactly solvable Gaussian setting, we prove that the excess risk is weighted by $\\Sigma_t$, and that both gradient descent and stochastic gradient descent inherit condition-number-dependent convergence. We then extend the "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We propose preconditioned flow matching, a precondition-then-match framework that transforms the target distribution into a more isotropic representation, trains the main flow in the transformed space, and maps generated samples back through the inverse transformation. We show theoretically that preconditioning reshapes the intermediate FM path and improves its conditioning.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that an effective preconditioner exists which can be applied without introducing new optimization difficulties or distorting the probability paths in a way that invalidates the flow matching objective.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Preconditioned flow matching transforms target distributions to isotropic representations to fix ill-conditioned optimization in standard flow matching.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Preconditioning transforms targets to isotropic space to reshape flow matching paths and fix ill-conditioned optimization.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"cd9b69194837fc6b9b94a2e208a38ebda53000ceb9da140991ee5027096165e7"},"source":{"id":"2603.02337","kind":"arxiv","version":2},"verdict":{"id":"49bc7cc3-4a02-4f18-b917-b759f5561df0","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T17:18:45.776522Z","strongest_claim":"We propose preconditioned flow matching, a precondition-then-match framework that transforms the target distribution into a more isotropic representation, trains the main flow in the transformed space, and maps generated samples back through the inverse transformation. We show theoretically that preconditioning reshapes the intermediate FM path and improves its conditioning.","one_line_summary":"Preconditioned flow matching transforms target distributions to isotropic representations to fix ill-conditioned optimization in standard flow matching.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that an effective preconditioner exists which can be applied without introducing new optimization difficulties or distorting the probability paths in a way that invalidates the flow matching objective.","pith_extraction_headline":"Preconditioning transforms targets to isotropic space to reshape flow matching paths and fix ill-conditioned optimization."},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"d9b1bff449fbb9c00d0394be33da87ca7dae6185863ec8da6aa0472ac33ca710"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"49bc7cc3-4a02-4f18-b917-b759f5561df0"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:09:23Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8aCW2Cmgi/xjEN3J6tnKkTRgICHFhtNNDhrqh/I0RTR6KwW6qpQqyDzr8iZbwfPT/PIvY7NLSjyAgFXTx4lEAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-22T23:31:05.229972Z"},"content_sha256":"f0ae283e5c392289625ccafb36c92b749349ef13d9b1fc8cdb89ad2056d8f166","schema_version":"1.0","event_id":"sha256:f0ae283e5c392289625ccafb36c92b749349ef13d9b1fc8cdb89ad2056d8f166"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP/bundle.json","state_url":"https://pith.science/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-22T23:31:05Z","links":{"resolver":"https://pith.science/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP","bundle":"https://pith.science/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP/bundle.json","state":"https://pith.science/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/SRQXSWF6VX2ACCUS64ZF2DZQBP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:SRQXSWF6VX2ACCUS64ZF2DZQBP","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"f82d9cbde890954c96bb660ae47ae3beffc3e8acf2003284ad1a1f3f4f704c72","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-02T19:09:15Z","title_canon_sha256":"e4433be4b722edab0843c62d68496fafab0d6be6ca150dacc1a83daa51daba46"},"schema_version":"1.0","source":{"id":"2603.02337","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2603.02337","created_at":"2026-05-18T03:09:23Z"},{"alias_kind":"arxiv_version","alias_value":"2603.02337v2","created_at":"2026-05-18T03:09:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.02337","created_at":"2026-05-18T03:09:23Z"},{"alias_kind":"pith_short_12","alias_value":"SRQXSWF6VX2A","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"SRQXSWF6VX2ACCUS","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"SRQXSWF6","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:f0ae283e5c392289625ccafb36c92b749349ef13d9b1fc8cdb89ad2056d8f166","target":"graph","created_at":"2026-05-18T03:09:23Z","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":"We propose preconditioned flow matching, a precondition-then-match framework that transforms the target distribution into a more isotropic representation, trains the main flow in the transformed space, and maps generated samples back through the inverse transformation. We show theoretically that preconditioning reshapes the intermediate FM path and improves its conditioning."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that an effective preconditioner exists which can be applied without introducing new optimization difficulties or distorting the probability paths in a way that invalidates the flow matching objective."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Preconditioned flow matching transforms target distributions to isotropic representations to fix ill-conditioned optimization in standard flow matching."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Preconditioning transforms targets to isotropic space to reshape flow matching paths and fix ill-conditioned optimization."}],"snapshot_sha256":"cd9b69194837fc6b9b94a2e208a38ebda53000ceb9da140991ee5027096165e7"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"d9b1bff449fbb9c00d0394be33da87ca7dae6185863ec8da6aa0472ac33ca710"},"paper":{"abstract_excerpt":"Flow matching (FM) learns vector fields by regressing stochastic velocity targets along intermediate distributions $p_t$. We identify a geometric optimization bottleneck in this regression problem: when the covariance $\\Sigma_t$ of $p_t$ is ill-conditioned, gradient-based training rapidly fits high-variance directions while making slow progress along low-variance ones. In an exactly solvable Gaussian setting, we prove that the excess risk is weighted by $\\Sigma_t$, and that both gradient descent and stochastic gradient descent inherit condition-number-dependent convergence. We then extend the ","authors_text":"Eldad Haber, Eshed Gal, Md Shahriar Rahim Siddiqui, Moshe Eliasof, Shadab Ahamed, Simon Ghyselincks","cross_cats":["cs.AI","cs.CV"],"headline":"Preconditioning transforms targets to isotropic space to reshape flow matching paths and fix ill-conditioned optimization.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-02T19:09:15Z","title":"Preconditioned Flow Matching"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.02337","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-15T17:18:45.776522Z","id":"49bc7cc3-4a02-4f18-b917-b759f5561df0","model_set":{"reader":"grok-4.3"},"one_line_summary":"Preconditioned flow matching transforms target distributions to isotropic representations to fix ill-conditioned optimization in standard flow matching.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Preconditioning transforms targets to isotropic space to reshape flow matching paths and fix ill-conditioned optimization.","strongest_claim":"We propose preconditioned flow matching, a precondition-then-match framework that transforms the target distribution into a more isotropic representation, trains the main flow in the transformed space, and maps generated samples back through the inverse transformation. We show theoretically that preconditioning reshapes the intermediate FM path and improves its conditioning.","weakest_assumption":"The assumption that an effective preconditioner exists which can be applied without introducing new optimization difficulties or distorting the probability paths in a way that invalidates the flow matching objective."}},"verdict_id":"49bc7cc3-4a02-4f18-b917-b759f5561df0"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:5bd665e45ea4c59699646aa546de887bbb84aee6fdcbf2dd78fb66fccf7e8ea3","target":"record","created_at":"2026-05-18T03:09:23Z","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":"f82d9cbde890954c96bb660ae47ae3beffc3e8acf2003284ad1a1f3f4f704c72","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-03-02T19:09:15Z","title_canon_sha256":"e4433be4b722edab0843c62d68496fafab0d6be6ca150dacc1a83daa51daba46"},"schema_version":"1.0","source":{"id":"2603.02337","kind":"arxiv","version":2}},"canonical_sha256":"94617958beadf4010a92f7325d0f300be322ebd7e4bd33b1db94139de227c6e5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"94617958beadf4010a92f7325d0f300be322ebd7e4bd33b1db94139de227c6e5","first_computed_at":"2026-05-18T03:09:23.166936Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:09:23.166936Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"N7jTCLWitaAYXplYzZAZ8ScEl3e/tMPATMfAqVlnBoMLcH7/nPO2N9XG0CooH5AuG4q4qf4Cghm34SwyDlraBA==","signature_status":"signed_v1","signed_at":"2026-05-18T03:09:23.167582Z","signed_message":"canonical_sha256_bytes"},"source_id":"2603.02337","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5bd665e45ea4c59699646aa546de887bbb84aee6fdcbf2dd78fb66fccf7e8ea3","sha256:f0ae283e5c392289625ccafb36c92b749349ef13d9b1fc8cdb89ad2056d8f166"],"state_sha256":"54a50d0c1b1fd2d647eadadeda3bdfc0610483b200c1e52334e29eec810b2465"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"gZV9KU9KbVRGWiVCuaFSQjHqbO0pbAAE56LCla5OkQycsZJGOkUPMIGZ5St+u8ePQY9NRsipw+Gdrpn6sOW4Aw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-22T23:31:05.234840Z","bundle_sha256":"57961808515a852c46f79e26869f96f8560de6710e402c08454811431a88eb8e"}}