{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:5WUKICFJOT6ELZVK5TCKUA7XBY","short_pith_number":"pith:5WUKICFJ","canonical_record":{"source":{"id":"2605.12754","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:07:13Z","cross_cats_sorted":[],"title_canon_sha256":"ec337b7f8d08f5718521ccb2d7e228c1894fa098ed2558d4f1aff1f4a0b201af","abstract_canon_sha256":"97ef234daf4cf06010e9ee6e9b9a32e3a78c958c7185b964b6d260f3bc939b12"},"schema_version":"1.0"},"canonical_sha256":"eda8a408a974fc45e6aaecc4aa03f70e29b7b577356ce7e4cb2e85887da32b41","source":{"kind":"arxiv","id":"2605.12754","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12754","created_at":"2026-05-18T03:09:48Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12754v1","created_at":"2026-05-18T03:09:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12754","created_at":"2026-05-18T03:09:48Z"},{"alias_kind":"pith_short_12","alias_value":"5WUKICFJOT6E","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"5WUKICFJOT6ELZVK","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"5WUKICFJ","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:5WUKICFJOT6ELZVK5TCKUA7XBY","target":"record","payload":{"canonical_record":{"source":{"id":"2605.12754","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:07:13Z","cross_cats_sorted":[],"title_canon_sha256":"ec337b7f8d08f5718521ccb2d7e228c1894fa098ed2558d4f1aff1f4a0b201af","abstract_canon_sha256":"97ef234daf4cf06010e9ee6e9b9a32e3a78c958c7185b964b6d260f3bc939b12"},"schema_version":"1.0"},"canonical_sha256":"eda8a408a974fc45e6aaecc4aa03f70e29b7b577356ce7e4cb2e85887da32b41","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:09:48.625831Z","signature_b64":"3/XWznsRhQvycH91IAzotBbGF7mJC3RQFhs3thZ4Obv2VPxnDgzWEbtiff0Sy30MVw4WOW+pYFMSAWmqMjATDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eda8a408a974fc45e6aaecc4aa03f70e29b7b577356ce7e4cb2e85887da32b41","last_reissued_at":"2026-05-18T03:09:48.625049Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:09:48.625049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.12754","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:09:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"u6gvWwthrVm20Pis4b/mv2Ew/ioBSyaeh2nVE1DYBGU0f2rbg5Z8lV3SljlU+4AaXQ15el0pFF3rJxGaFWbVDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T10:24:35.055453Z"},"content_sha256":"b1270d73a246a84afbac2316ac5c8302661f83ad2cd0e38295c44686e597bd28","schema_version":"1.0","event_id":"sha256:b1270d73a246a84afbac2316ac5c8302661f83ad2cd0e38295c44686e597bd28"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:5WUKICFJOT6ELZVK5TCKUA7XBY","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Constraint-Aware Flow Matching: Decision Aligned End-to-End Training for Constrained Sampling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Constraint-Aware Flow Matching trains generative models by embedding constraint projections directly into the flow objective, closing the training-sampling mismatch that degrades quality in existing methods.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Ferdinando Fioretto, Jacob K. Christopher, James E. Warner","submitted_at":"2026-05-12T21:07:13Z","abstract_excerpt":"Deep generative models provide state-of-the-art performance across a wide array of applications, with recent studies showing increasing applicability for science and engineering. Despite a growing corpus of literature focused on the integration of physics-based constraints into the generation process, existing approaches fail to enforce strict constraint satisfaction while maintaining sample quality. In particular, training-free constrained sampling methods, while providing per-sample feasibility guarantees, introduce a fundamental mismatch between the training objective and the constrained sa"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"By aligning the model's learned dynamics with the constrained sampling process, the proposed method mitigates distributional shift induced by projection-based corrections, enabling high-quality constrained generation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That folding constraint projections into the training objective will produce stable gradients and will not create new optimization difficulties or unintended biases in the learned distribution.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Constraint-Aware Flow Matching integrates constraint projections into the flow matching training objective to align model dynamics with constrained sampling and reduce distributional shift.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Constraint-Aware Flow Matching trains generative models by embedding constraint projections directly into the flow objective, closing the training-sampling mismatch that degrades quality in existing methods.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c03685509711668dd5f0d6b1bdc44cf106efd6fdc4ae0807cdf5342a1a5450e9"},"source":{"id":"2605.12754","kind":"arxiv","version":1},"verdict":{"id":"d59e01df-7faa-40d2-b392-82685ced3c03","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T21:13:24.112789Z","strongest_claim":"By aligning the model's learned dynamics with the constrained sampling process, the proposed method mitigates distributional shift induced by projection-based corrections, enabling high-quality constrained generation.","one_line_summary":"Constraint-Aware Flow Matching integrates constraint projections into the flow matching training objective to align model dynamics with constrained sampling and reduce distributional shift.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That folding constraint projections into the training objective will produce stable gradients and will not create new optimization difficulties or unintended biases in the learned distribution.","pith_extraction_headline":"Constraint-Aware Flow Matching trains generative models by embedding constraint projections directly into the flow objective, closing the training-sampling mismatch that degrades quality in existing methods."},"references":{"count":59,"sample":[{"doi":"","year":2022,"title":"High-resolution image synthesis with latent diffusion models","work_id":"3cba96fb-e636-4639-8d43-2f25ce21d4d1","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Flowception: Temporally expansive flow matching for video generation.arXiv preprint arXiv:2512.11438, 2025","work_id":"3a9b504d-a19e-4a07-ba64-1018e56764d5","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Pyramidal flow matching for efficient video generative modeling","work_id":"86ba3b09-fc7d-4d26-acce-10677e0b5d69","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Diffusebot: Breeding soft robots with physics-augmented generative diffusion models.Advances in Neural Information Processing Systems, 36:44398–44423, 2023","work_id":"0905b2d8-8a9a-465d-9f09-165ba6793597","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Con- strained diffusion for accelerated structure relaxation of inorganic solids with point defects.arXiv preprint arXiv:2602.19153, 2026","work_id":"171d141e-c72e-405c-adf8-0fcb8e13b423","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":59,"snapshot_sha256":"c347ae6f8f348f162f29cf305a46abca77dcc6119cf1e6f13990bcc47b1a457b","internal_anchors":5},"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"},"verdict_id":"d59e01df-7faa-40d2-b392-82685ced3c03"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T03:09:48Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"fAgkUw0DbwrxoblQoBPMOYwREBQBw2dGDHlGGpeQa2084z6h6038CbiZezdEY/NogEkc60nQl7hpjT9ZQOjqCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-15T10:24:35.057008Z"},"content_sha256":"e93b5b517aec8b0cc38ff7d4800054c274f3cb1ade117a2a816c06b192a95f25","schema_version":"1.0","event_id":"sha256:e93b5b517aec8b0cc38ff7d4800054c274f3cb1ade117a2a816c06b192a95f25"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/5WUKICFJOT6ELZVK5TCKUA7XBY/bundle.json","state_url":"https://pith.science/pith/5WUKICFJOT6ELZVK5TCKUA7XBY/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/5WUKICFJOT6ELZVK5TCKUA7XBY/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":"2026-08-15T10:24:35Z","links":{"resolver":"https://pith.science/pith/5WUKICFJOT6ELZVK5TCKUA7XBY","bundle":"https://pith.science/pith/5WUKICFJOT6ELZVK5TCKUA7XBY/bundle.json","state":"https://pith.science/pith/5WUKICFJOT6ELZVK5TCKUA7XBY/state.json","well_known_bundle":"https://pith.science/.well-known/pith/5WUKICFJOT6ELZVK5TCKUA7XBY/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:5WUKICFJOT6ELZVK5TCKUA7XBY","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":"97ef234daf4cf06010e9ee6e9b9a32e3a78c958c7185b964b6d260f3bc939b12","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:07:13Z","title_canon_sha256":"ec337b7f8d08f5718521ccb2d7e228c1894fa098ed2558d4f1aff1f4a0b201af"},"schema_version":"1.0","source":{"id":"2605.12754","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.12754","created_at":"2026-05-18T03:09:48Z"},{"alias_kind":"arxiv_version","alias_value":"2605.12754v1","created_at":"2026-05-18T03:09:48Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.12754","created_at":"2026-05-18T03:09:48Z"},{"alias_kind":"pith_short_12","alias_value":"5WUKICFJOT6E","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"5WUKICFJOT6ELZVK","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"5WUKICFJ","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:e93b5b517aec8b0cc38ff7d4800054c274f3cb1ade117a2a816c06b192a95f25","target":"graph","created_at":"2026-05-18T03:09:48Z","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 aligning the model's learned dynamics with the constrained sampling process, the proposed method mitigates distributional shift induced by projection-based corrections, enabling high-quality constrained generation."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That folding constraint projections into the training objective will produce stable gradients and will not create new optimization difficulties or unintended biases in the learned distribution."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Constraint-Aware Flow Matching integrates constraint projections into the flow matching training objective to align model dynamics with constrained sampling and reduce distributional shift."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Constraint-Aware Flow Matching trains generative models by embedding constraint projections directly into the flow objective, closing the training-sampling mismatch that degrades quality in existing methods."}],"snapshot_sha256":"c03685509711668dd5f0d6b1bdc44cf106efd6fdc4ae0807cdf5342a1a5450e9"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Deep generative models provide state-of-the-art performance across a wide array of applications, with recent studies showing increasing applicability for science and engineering. Despite a growing corpus of literature focused on the integration of physics-based constraints into the generation process, existing approaches fail to enforce strict constraint satisfaction while maintaining sample quality. In particular, training-free constrained sampling methods, while providing per-sample feasibility guarantees, introduce a fundamental mismatch between the training objective and the constrained sa","authors_text":"Ferdinando Fioretto, Jacob K. Christopher, James E. Warner","cross_cats":[],"headline":"Constraint-Aware Flow Matching trains generative models by embedding constraint projections directly into the flow objective, closing the training-sampling mismatch that degrades quality in existing methods.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:07:13Z","title":"Constraint-Aware Flow Matching: Decision Aligned End-to-End Training for Constrained Sampling"},"references":{"count":59,"internal_anchors":5,"resolved_work":59,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"High-resolution image synthesis with latent diffusion models","work_id":"3cba96fb-e636-4639-8d43-2f25ce21d4d1","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Flowception: Temporally expansive flow matching for video generation.arXiv preprint arXiv:2512.11438, 2025","work_id":"3a9b504d-a19e-4a07-ba64-1018e56764d5","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Pyramidal flow matching for efficient video generative modeling","work_id":"86ba3b09-fc7d-4d26-acce-10677e0b5d69","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Diffusebot: Breeding soft robots with physics-augmented generative diffusion models.Advances in Neural Information Processing Systems, 36:44398–44423, 2023","work_id":"0905b2d8-8a9a-465d-9f09-165ba6793597","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Con- strained diffusion for accelerated structure relaxation of inorganic solids with point defects.arXiv preprint arXiv:2602.19153, 2026","work_id":"171d141e-c72e-405c-adf8-0fcb8e13b423","year":2026}],"snapshot_sha256":"c347ae6f8f348f162f29cf305a46abca77dcc6119cf1e6f13990bcc47b1a457b"},"source":{"id":"2605.12754","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T21:13:24.112789Z","id":"d59e01df-7faa-40d2-b392-82685ced3c03","model_set":{"reader":"grok-4.3"},"one_line_summary":"Constraint-Aware Flow Matching integrates constraint projections into the flow matching training objective to align model dynamics with constrained sampling and reduce distributional shift.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Constraint-Aware Flow Matching trains generative models by embedding constraint projections directly into the flow objective, closing the training-sampling mismatch that degrades quality in existing methods.","strongest_claim":"By aligning the model's learned dynamics with the constrained sampling process, the proposed method mitigates distributional shift induced by projection-based corrections, enabling high-quality constrained generation.","weakest_assumption":"That folding constraint projections into the training objective will produce stable gradients and will not create new optimization difficulties or unintended biases in the learned distribution."}},"verdict_id":"d59e01df-7faa-40d2-b392-82685ced3c03"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b1270d73a246a84afbac2316ac5c8302661f83ad2cd0e38295c44686e597bd28","target":"record","created_at":"2026-05-18T03:09:48Z","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":"97ef234daf4cf06010e9ee6e9b9a32e3a78c958c7185b964b6d260f3bc939b12","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T21:07:13Z","title_canon_sha256":"ec337b7f8d08f5718521ccb2d7e228c1894fa098ed2558d4f1aff1f4a0b201af"},"schema_version":"1.0","source":{"id":"2605.12754","kind":"arxiv","version":1}},"canonical_sha256":"eda8a408a974fc45e6aaecc4aa03f70e29b7b577356ce7e4cb2e85887da32b41","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"eda8a408a974fc45e6aaecc4aa03f70e29b7b577356ce7e4cb2e85887da32b41","first_computed_at":"2026-05-18T03:09:48.625049Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:09:48.625049Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"3/XWznsRhQvycH91IAzotBbGF7mJC3RQFhs3thZ4Obv2VPxnDgzWEbtiff0Sy30MVw4WOW+pYFMSAWmqMjATDQ==","signature_status":"signed_v1","signed_at":"2026-05-18T03:09:48.625831Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.12754","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b1270d73a246a84afbac2316ac5c8302661f83ad2cd0e38295c44686e597bd28","sha256:e93b5b517aec8b0cc38ff7d4800054c274f3cb1ade117a2a816c06b192a95f25"],"state_sha256":"a11c39344c91fb4176b3db0bf774585ecdc53fb430b513a46d23e9fd4375e31e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"NLAU/+8/X5TPIMdogVAJMzfT9OQodxXTXLgZjLhYGK1mu/yfocESyc5T3wgMtoB8HtJxriHgDrWN+ox0FMP/Bg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-15T10:24:35.062102Z","bundle_sha256":"a342304d8156d8882b41c367d64191fff74e947640af294d718663b3846d4989"}}