{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:R7BCWP37ECGZDXQZV7SQJQGYXO","short_pith_number":"pith:R7BCWP37","canonical_record":{"source":{"id":"2605.01361","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T10:10:16Z","cross_cats_sorted":[],"title_canon_sha256":"852e7d258e107a38363639d4c2a3d5387880a81c75ce0a7657c7959e785f24c2","abstract_canon_sha256":"0e1da7c7205cc484b24c83ffa9287baeea543716dce7292a5c5abe5b93cdba58"},"schema_version":"1.0"},"canonical_sha256":"8fc22b3f7f208d91de19afe504c0d8bb9fb69bd50eb5db9db3a4ea42446552c5","source":{"kind":"arxiv","id":"2605.01361","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01361","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01361v2","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01361","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"pith_short_12","alias_value":"R7BCWP37ECGZ","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"pith_short_16","alias_value":"R7BCWP37ECGZDXQZ","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"pith_short_8","alias_value":"R7BCWP37","created_at":"2026-05-20T01:06:09Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:R7BCWP37ECGZDXQZV7SQJQGYXO","target":"record","payload":{"canonical_record":{"source":{"id":"2605.01361","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T10:10:16Z","cross_cats_sorted":[],"title_canon_sha256":"852e7d258e107a38363639d4c2a3d5387880a81c75ce0a7657c7959e785f24c2","abstract_canon_sha256":"0e1da7c7205cc484b24c83ffa9287baeea543716dce7292a5c5abe5b93cdba58"},"schema_version":"1.0"},"canonical_sha256":"8fc22b3f7f208d91de19afe504c0d8bb9fb69bd50eb5db9db3a4ea42446552c5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:06:09.576592Z","signature_b64":"sd1wVDUmMqx0ik8ZPs0bMrG8yQb+UkJA4MPhFxzMs4CCNerxolxr+5vrfi0j3RNJ9Vv3b7hVB2wDxsrHaNdDAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8fc22b3f7f208d91de19afe504c0d8bb9fb69bd50eb5db9db3a4ea42446552c5","last_reissued_at":"2026-05-20T01:06:09.575779Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:06:09.575779Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.01361","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":"2026-05-20T01:06:09Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/JMrajsofbtDYNs7TipwOu67ovhbHOtLkiesslZwLhXYGzA+f/kJWQy8xmPD/qEf49c9BkWKwVAE3IgmJcz5Bw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-30T16:48:03.429967Z"},"content_sha256":"5c77c854455dc78c3fe9f729401192b2d8cb781761597de276d062cdf035277b","schema_version":"1.0","event_id":"sha256:5c77c854455dc78c3fe9f729401192b2d8cb781761597de276d062cdf035277b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:R7BCWP37ECGZDXQZV7SQJQGYXO","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Decision-Focused Learning via Tangent-Space Projection of Prediction Error","license":"http://creativecommons.org/licenses/by/4.0/","headline":"The regret gradient equals the prediction error projected onto the tangent space of active constraints and scaled by local curvature.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Junhyeong Lee, Sangjin Jin, Yongjae Lee","submitted_at":"2026-05-02T10:10:16Z","abstract_excerpt":"Decision-Focused Learning (DFL) trains predictors to improve downstream decision quality, but computing regret gradients typically requires differentiating through solvers or relying on surrogate losses, which can be computationally expensive or deviate from the true objective. We show that, under standard regularity with locally stable active constraints, the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature. This reveals that regret gradients can be obtained by "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"under standard regularity with locally stable active constraints","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Regret gradients in DFL are the tangent-space projection of prediction error scaled by curvature, enabling efficient direct computation without differentiating through solvers.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"The regret gradient equals the prediction error projected onto the tangent space of active constraints and scaled by local curvature.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"c68f90176199b129e7f1bd4a7a71aa7502db232d46e07ce83a5a8bc1a1d14fd7"},"source":{"id":"2605.01361","kind":"arxiv","version":2},"verdict":{"id":"e24501b6-d301-423c-8396-0f09b39f9b4d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-09T14:14:05.985494Z","strongest_claim":"the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature","one_line_summary":"Regret gradients in DFL are the tangent-space projection of prediction error scaled by curvature, enabling efficient direct computation without differentiating through solvers.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"under standard regularity with locally stable active constraints","pith_extraction_headline":"The regret gradient equals the prediction error projected onto the tangent space of active constraints and scaled by local curvature."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.01361/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T17:19:59.017217Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4275449424a338c426eba6ec227a27d5ea1594f2d22198a853c5c6bde9cca3bc"},"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"},"verdict_id":"e24501b6-d301-423c-8396-0f09b39f9b4d"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T01:06:09Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"VLMzj0AY/pbOA4kvf4UXmZjdCHspzMpkmKHBYdfAdnSHYEuKAw6ka5lEzM5eVydh4QWkyw10cb309mShZ5ulAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-30T16:48:03.430465Z"},"content_sha256":"2bd1783b923628232720322aef967684097bb8c78974aa92638d281f3d0499a2","schema_version":"1.0","event_id":"sha256:2bd1783b923628232720322aef967684097bb8c78974aa92638d281f3d0499a2"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/R7BCWP37ECGZDXQZV7SQJQGYXO/bundle.json","state_url":"https://pith.science/pith/R7BCWP37ECGZDXQZV7SQJQGYXO/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/R7BCWP37ECGZDXQZV7SQJQGYXO/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":"2026-05-30T16:48:03Z","links":{"resolver":"https://pith.science/pith/R7BCWP37ECGZDXQZV7SQJQGYXO","bundle":"https://pith.science/pith/R7BCWP37ECGZDXQZV7SQJQGYXO/bundle.json","state":"https://pith.science/pith/R7BCWP37ECGZDXQZV7SQJQGYXO/state.json","well_known_bundle":"https://pith.science/.well-known/pith/R7BCWP37ECGZDXQZV7SQJQGYXO/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:R7BCWP37ECGZDXQZV7SQJQGYXO","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":"0e1da7c7205cc484b24c83ffa9287baeea543716dce7292a5c5abe5b93cdba58","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T10:10:16Z","title_canon_sha256":"852e7d258e107a38363639d4c2a3d5387880a81c75ce0a7657c7959e785f24c2"},"schema_version":"1.0","source":{"id":"2605.01361","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01361","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01361v2","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01361","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"pith_short_12","alias_value":"R7BCWP37ECGZ","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"pith_short_16","alias_value":"R7BCWP37ECGZDXQZ","created_at":"2026-05-20T01:06:09Z"},{"alias_kind":"pith_short_8","alias_value":"R7BCWP37","created_at":"2026-05-20T01:06:09Z"}],"graph_snapshots":[{"event_id":"sha256:2bd1783b923628232720322aef967684097bb8c78974aa92638d281f3d0499a2","target":"graph","created_at":"2026-05-20T01:06:09Z","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":"the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature"},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"under standard regularity with locally stable active constraints"},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Regret gradients in DFL are the tangent-space projection of prediction error scaled by curvature, enabling efficient direct computation without differentiating through solvers."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"The regret gradient equals the prediction error projected onto the tangent space of active constraints and scaled by local curvature."}],"snapshot_sha256":"c68f90176199b129e7f1bd4a7a71aa7502db232d46e07ce83a5a8bc1a1d14fd7"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T17:19:59.017217Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.01361/integrity.json","findings":[],"snapshot_sha256":"4275449424a338c426eba6ec227a27d5ea1594f2d22198a853c5c6bde9cca3bc","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Decision-Focused Learning (DFL) trains predictors to improve downstream decision quality, but computing regret gradients typically requires differentiating through solvers or relying on surrogate losses, which can be computationally expensive or deviate from the true objective. We show that, under standard regularity with locally stable active constraints, the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature. This reveals that regret gradients can be obtained by ","authors_text":"Junhyeong Lee, Sangjin Jin, Yongjae Lee","cross_cats":[],"headline":"The regret gradient equals the prediction error projected onto the tangent space of active constraints and scaled by local curvature.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T10:10:16Z","title":"Decision-Focused Learning via Tangent-Space Projection of Prediction Error"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.01361","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-09T14:14:05.985494Z","id":"e24501b6-d301-423c-8396-0f09b39f9b4d","model_set":{"reader":"grok-4.3"},"one_line_summary":"Regret gradients in DFL are the tangent-space projection of prediction error scaled by curvature, enabling efficient direct computation without differentiating through solvers.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"The regret gradient equals the prediction error projected onto the tangent space of active constraints and scaled by local curvature.","strongest_claim":"the regret gradient admits a closed-form geometric characterization, equivalent to the prediction error projected onto the tangent space of active constraints, scaled by local curvature","weakest_assumption":"under standard regularity with locally stable active constraints"}},"verdict_id":"e24501b6-d301-423c-8396-0f09b39f9b4d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:5c77c854455dc78c3fe9f729401192b2d8cb781761597de276d062cdf035277b","target":"record","created_at":"2026-05-20T01:06:09Z","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":"0e1da7c7205cc484b24c83ffa9287baeea543716dce7292a5c5abe5b93cdba58","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T10:10:16Z","title_canon_sha256":"852e7d258e107a38363639d4c2a3d5387880a81c75ce0a7657c7959e785f24c2"},"schema_version":"1.0","source":{"id":"2605.01361","kind":"arxiv","version":2}},"canonical_sha256":"8fc22b3f7f208d91de19afe504c0d8bb9fb69bd50eb5db9db3a4ea42446552c5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8fc22b3f7f208d91de19afe504c0d8bb9fb69bd50eb5db9db3a4ea42446552c5","first_computed_at":"2026-05-20T01:06:09.575779Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T01:06:09.575779Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"sd1wVDUmMqx0ik8ZPs0bMrG8yQb+UkJA4MPhFxzMs4CCNerxolxr+5vrfi0j3RNJ9Vv3b7hVB2wDxsrHaNdDAg==","signature_status":"signed_v1","signed_at":"2026-05-20T01:06:09.576592Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.01361","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5c77c854455dc78c3fe9f729401192b2d8cb781761597de276d062cdf035277b","sha256:2bd1783b923628232720322aef967684097bb8c78974aa92638d281f3d0499a2"],"state_sha256":"572294833576170e932c9ddad01a019b806049b71f285d393dbde9e21c7ef06b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"HzBHNNN3s1/wvL6/ClrtjgspsdOKLjh7dhRVGH3fj7Bulyp2JlMlkPt7HUUs+jXZGAW6wiFwa/vmvZSRtaxLDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-30T16:48:03.434611Z","bundle_sha256":"8dae91300f16586ddb97e8c3366a0ce606976a053ef9040dfd867ec8536e9522"}}