{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:HRFP2HBBY3JQ4IYM6BUYUXZP5B","short_pith_number":"pith:HRFP2HBB","canonical_record":{"source":{"id":"2605.02167","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T02:56:29Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"b2fb7701d7f6943f2b83bc3d1e0ea6eefa04b5553b9e6ff82a503821bc002610","abstract_canon_sha256":"40716b17f1de3f5223e39557b49855681e82e2a5d7f34e0887b541cf828854e7"},"schema_version":"1.0"},"canonical_sha256":"3c4afd1c21c6d30e230cf0698a5f2fe87d3b7912208e81c3fdc38be6ebf49322","source":{"kind":"arxiv","id":"2605.02167","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.02167","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"arxiv_version","alias_value":"2605.02167v3","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.02167","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_12","alias_value":"HRFP2HBBY3JQ","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_16","alias_value":"HRFP2HBBY3JQ4IYM","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_8","alias_value":"HRFP2HBB","created_at":"2026-05-20T00:03:13Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:HRFP2HBBY3JQ4IYM6BUYUXZP5B","target":"record","payload":{"canonical_record":{"source":{"id":"2605.02167","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T02:56:29Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"b2fb7701d7f6943f2b83bc3d1e0ea6eefa04b5553b9e6ff82a503821bc002610","abstract_canon_sha256":"40716b17f1de3f5223e39557b49855681e82e2a5d7f34e0887b541cf828854e7"},"schema_version":"1.0"},"canonical_sha256":"3c4afd1c21c6d30e230cf0698a5f2fe87d3b7912208e81c3fdc38be6ebf49322","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:03:13.717281Z","signature_b64":"ykMPkiiLVH260eWwmRUZxE5tZz+kqK7phzr5wy5nHqkaFaGRUlhy5Lt3q5d9/LA8eI0lnmC9Idw67qKaMwrQCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3c4afd1c21c6d30e230cf0698a5f2fe87d3b7912208e81c3fdc38be6ebf49322","last_reissued_at":"2026-05-20T00:03:13.716203Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:03:13.716203Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.02167","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"xJzRLoHEUuWVyDFvu4ST5fJcnf03t7rp1MqJYY/5CHdIE6bbRMOWVnmkGSEDvxbY/C/AmYv59Rm0Qb0pRHTEBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-21T16:08:11.284842Z"},"content_sha256":"94678ad2141f371ebea6db9c3f60fc69ea2ffb351dbfc2779072aa3b30e6d33a","schema_version":"1.0","event_id":"sha256:94678ad2141f371ebea6db9c3f60fc69ea2ffb351dbfc2779072aa3b30e6d33a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:HRFP2HBBY3JQ4IYM6BUYUXZP5B","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Manifold-Aligned Guided Integrated Gradients for Reliable Feature Attribution","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"By constructing attribution paths in a variational autoencoder's latent space, MA-GIG produces more faithful feature attributions than standard path-based methods.","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Jaesik Choi, Kyowoon Lee, Seongwoo Lim, Soyeon Kim","submitted_at":"2026-05-04T02:56:29Z","abstract_excerpt":"Feature attribution is central to diagnosing and trusting deep neural networks, and Integrated Gradients (IG) is widely used due to its axiomatic properties. However, IG can yield unreliable explanations when the integration path between a baseline and the input passes through regions with noisy gradients. While Guided Integrated Gradients reduces this sensitivity by adaptively updating low-gradient-magnitude features, input-space guidance still produces intermediate inputs that deviate from the data manifold. To address this limitation, we propose \\emph{Manifold-Aligned Guided Integrated Grad"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Through qualitative and quantitative evaluations, we demonstrate that MA-GIG produces faithful explanations by aggregating gradients on path features proximal to the input. Consequently, our method reduces off-manifold noise and outperforms prior path-based attribution methods across multiple datasets and classifiers.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"A pre-trained variational autoencoder accurately captures the data manifold, and decoded latent-space paths therefore yield gradient aggregations that are more faithful than those obtained from input-space paths.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MA-GIG improves Integrated Gradients by performing path integration in the latent space of a pre-trained VAE so that decoded points remain closer to the learned data manifold and reduce off-manifold gradient noise.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"By constructing attribution paths in a variational autoencoder's latent space, MA-GIG produces more faithful feature attributions than standard path-based methods.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"da49e77c1ae024df8bbfcd233ccf74667ae1c8a6129c2b8756eb8f220c9d74b7"},"source":{"id":"2605.02167","kind":"arxiv","version":3},"verdict":{"id":"a49161b5-9e9f-4bf8-9697-096271874411","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T19:04:06.763098Z","strongest_claim":"Through qualitative and quantitative evaluations, we demonstrate that MA-GIG produces faithful explanations by aggregating gradients on path features proximal to the input. Consequently, our method reduces off-manifold noise and outperforms prior path-based attribution methods across multiple datasets and classifiers.","one_line_summary":"MA-GIG improves Integrated Gradients by performing path integration in the latent space of a pre-trained VAE so that decoded points remain closer to the learned data manifold and reduce off-manifold gradient noise.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"A pre-trained variational autoencoder accurately captures the data manifold, and decoded latent-space paths therefore yield gradient aggregations that are more faithful than those obtained from input-space paths.","pith_extraction_headline":"By constructing attribution paths in a variational autoencoder's latent space, MA-GIG produces more faithful feature attributions than standard path-based methods."},"integrity":{"clean":false,"summary":{"advisory":1,"critical":0,"by_detector":{"doi_compliance":{"total":1,"advisory":1,"critical":0,"informational":0}},"informational":0},"endpoint":"/pith/2605.02167/integrity.json","findings":[{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TPAMI.2024.3388092.12) was visible in the surrounding text but could not be confirmed against doi.org as printed.","detector":"doi_compliance","severity":"advisory","ref_index":10,"audited_at":"2026-05-19T16:37:40.734851Z","detected_doi":"10.1109/TPAMI.2024.3388092.12","finding_type":"recoverable_identifier","verdict_class":"incontrovertible","detected_arxiv_id":null}],"available":true,"detectors_run":[{"name":"doi_compliance","ran_at":"2026-05-19T16:37:40.734851Z","status":"completed","version":"1.0.0","findings_count":1}],"snapshot_sha256":"6fccd978da247c4870a3c51b5dbb3efe0cfbd23f452b85d2aab9011de3bf0b27"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"30f785f5267aa9bde0dcaece522764377f27430f389a61e376b427a8d0bf7160"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"a49161b5-9e9f-4bf8-9697-096271874411"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:03:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"BDPSkyz+XNoJ+lazkqcrVPMrM0Gq4jsFD3VRuLU5ELJW2bPUK+e/ViU4v7SJ9F+8HOLUv4mvHjNzMLr+CGUeAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-21T16:08:11.285423Z"},"content_sha256":"13d4eb732766259573c9813e1b36b5f42f1c366568e5fb3e3990a80baaaea866","schema_version":"1.0","event_id":"sha256:13d4eb732766259573c9813e1b36b5f42f1c366568e5fb3e3990a80baaaea866"},{"event_type":"integrity_finding","subject_pith_number":"pith:2026:HRFP2HBBY3JQ4IYM6BUYUXZP5B","target":"integrity","payload":{"note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TPAMI.2024.3388092.12) was visible in the surrounding text but could not be confirmed against doi.org as printed.","snippet":"ISSN 1939-3539. doi: 10.1109/TPAMI.2024. 3388092. 12 Manifold-Aligned Guided Integrated Gradients for Reliable Feature Attribution Appendix Table of Contents Appendix A.Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .","arxiv_id":"2605.02167","detector":"doi_compliance","evidence":{"ref_index":10,"verdict_class":"incontrovertible","resolved_title":null,"printed_excerpt":"ISSN 1939-3539. doi: 10.1109/TPAMI.2024. 3388092. 12 Manifold-Aligned Guided Integrated Gradients for Reliable Feature Attribution Appendix Table of Contents Appendix A.Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .","reconstructed_doi":"10.1109/TPAMI.2024.3388092.12"},"severity":"advisory","ref_index":10,"audited_at":"2026-05-19T16:37:40.734851Z","event_type":"pith.integrity.v1","detected_doi":"10.1109/TPAMI.2024.3388092.12","detector_url":"https://pith.science/pith-integrity-protocol#doi_compliance","external_url":null,"finding_type":"recoverable_identifier","evidence_hash":"bea8720c4e3987270f5b1edb3ae6a8d046f1e96fbc5dc55142ca5ffa08dbe7db","paper_version":2,"verdict_class":"incontrovertible","resolved_title":null,"detector_version":"1.0.0","detected_arxiv_id":null,"integrity_event_id":2070,"payload_sha256":"b7153744c35265329e0695cb02f763bbb462c30380f92f59e433b58f10235f12","signature_b64":"157i9/qHwYXv9kDVdvSYjB577Fk687aY/zLvZh1kpr7VMuTH8Bqm96QouecBTq/HwGyStyiuPm9SOPL77aQGAQ==","signing_key_id":"pith-v1-2026-05"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-19T16:42:13Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"U7BUoa50w17wdLn9WsJw8t8GiO4w9BC8EptNoMw9IGFwkTvP4W8/mVAooznDy7LgIHa24fvyQZ/c21AHw94yBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-21T16:08:11.286378Z"},"content_sha256":"835ecadef92dd3b028a3ca7290f3fa83d9b212a41ac8216fde577cffe234bf81","schema_version":"1.0","event_id":"sha256:835ecadef92dd3b028a3ca7290f3fa83d9b212a41ac8216fde577cffe234bf81"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B/bundle.json","state_url":"https://pith.science/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B/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":"2026-05-21T16:08:11Z","links":{"resolver":"https://pith.science/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B","bundle":"https://pith.science/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B/bundle.json","state":"https://pith.science/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B/state.json","well_known_bundle":"https://pith.science/.well-known/pith/HRFP2HBBY3JQ4IYM6BUYUXZP5B/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HRFP2HBBY3JQ4IYM6BUYUXZP5B","merge_version":"pith-open-graph-merge-v1","event_count":3,"valid_event_count":3,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"40716b17f1de3f5223e39557b49855681e82e2a5d7f34e0887b541cf828854e7","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T02:56:29Z","title_canon_sha256":"b2fb7701d7f6943f2b83bc3d1e0ea6eefa04b5553b9e6ff82a503821bc002610"},"schema_version":"1.0","source":{"id":"2605.02167","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.02167","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"arxiv_version","alias_value":"2605.02167v3","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.02167","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_12","alias_value":"HRFP2HBBY3JQ","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_16","alias_value":"HRFP2HBBY3JQ4IYM","created_at":"2026-05-20T00:03:13Z"},{"alias_kind":"pith_short_8","alias_value":"HRFP2HBB","created_at":"2026-05-20T00:03:13Z"}],"graph_snapshots":[{"event_id":"sha256:13d4eb732766259573c9813e1b36b5f42f1c366568e5fb3e3990a80baaaea866","target":"graph","created_at":"2026-05-20T00:03:13Z","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":"Through qualitative and quantitative evaluations, we demonstrate that MA-GIG produces faithful explanations by aggregating gradients on path features proximal to the input. Consequently, our method reduces off-manifold noise and outperforms prior path-based attribution methods across multiple datasets and classifiers."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"A pre-trained variational autoencoder accurately captures the data manifold, and decoded latent-space paths therefore yield gradient aggregations that are more faithful than those obtained from input-space paths."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"MA-GIG improves Integrated Gradients by performing path integration in the latent space of a pre-trained VAE so that decoded points remain closer to the learned data manifold and reduce off-manifold gradient noise."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"By constructing attribution paths in a variational autoencoder's latent space, MA-GIG produces more faithful feature attributions than standard path-based methods."}],"snapshot_sha256":"da49e77c1ae024df8bbfcd233ccf74667ae1c8a6129c2b8756eb8f220c9d74b7"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"30f785f5267aa9bde0dcaece522764377f27430f389a61e376b427a8d0bf7160"},"integrity":{"available":true,"clean":false,"detectors_run":[{"findings_count":1,"name":"doi_compliance","ran_at":"2026-05-19T16:37:40.734851Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.02167/integrity.json","findings":[{"audited_at":"2026-05-19T16:37:40.734851Z","detected_arxiv_id":null,"detected_doi":"10.1109/TPAMI.2024.3388092.12","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TPAMI.2024.3388092.12) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":10,"severity":"advisory","verdict_class":"incontrovertible"}],"snapshot_sha256":"6fccd978da247c4870a3c51b5dbb3efe0cfbd23f452b85d2aab9011de3bf0b27","summary":{"advisory":1,"by_detector":{"doi_compliance":{"advisory":1,"critical":0,"informational":0,"total":1}},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Feature attribution is central to diagnosing and trusting deep neural networks, and Integrated Gradients (IG) is widely used due to its axiomatic properties. However, IG can yield unreliable explanations when the integration path between a baseline and the input passes through regions with noisy gradients. While Guided Integrated Gradients reduces this sensitivity by adaptively updating low-gradient-magnitude features, input-space guidance still produces intermediate inputs that deviate from the data manifold. To address this limitation, we propose \\emph{Manifold-Aligned Guided Integrated Grad","authors_text":"Jaesik Choi, Kyowoon Lee, Seongwoo Lim, Soyeon Kim","cross_cats":["cs.AI","cs.CV"],"headline":"By constructing attribution paths in a variational autoencoder's latent space, MA-GIG produces more faithful feature attributions than standard path-based methods.","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T02:56:29Z","title":"Manifold-Aligned Guided Integrated Gradients for Reliable Feature Attribution"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.02167","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-08T19:04:06.763098Z","id":"a49161b5-9e9f-4bf8-9697-096271874411","model_set":{"reader":"grok-4.3"},"one_line_summary":"MA-GIG improves Integrated Gradients by performing path integration in the latent space of a pre-trained VAE so that decoded points remain closer to the learned data manifold and reduce off-manifold gradient noise.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"By constructing attribution paths in a variational autoencoder's latent space, MA-GIG produces more faithful feature attributions than standard path-based methods.","strongest_claim":"Through qualitative and quantitative evaluations, we demonstrate that MA-GIG produces faithful explanations by aggregating gradients on path features proximal to the input. Consequently, our method reduces off-manifold noise and outperforms prior path-based attribution methods across multiple datasets and classifiers.","weakest_assumption":"A pre-trained variational autoencoder accurately captures the data manifold, and decoded latent-space paths therefore yield gradient aggregations that are more faithful than those obtained from input-space paths."}},"verdict_id":"a49161b5-9e9f-4bf8-9697-096271874411"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:94678ad2141f371ebea6db9c3f60fc69ea2ffb351dbfc2779072aa3b30e6d33a","target":"record","created_at":"2026-05-20T00:03:13Z","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":"40716b17f1de3f5223e39557b49855681e82e2a5d7f34e0887b541cf828854e7","cross_cats_sorted":["cs.AI","cs.CV"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-04T02:56:29Z","title_canon_sha256":"b2fb7701d7f6943f2b83bc3d1e0ea6eefa04b5553b9e6ff82a503821bc002610"},"schema_version":"1.0","source":{"id":"2605.02167","kind":"arxiv","version":3}},"canonical_sha256":"3c4afd1c21c6d30e230cf0698a5f2fe87d3b7912208e81c3fdc38be6ebf49322","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3c4afd1c21c6d30e230cf0698a5f2fe87d3b7912208e81c3fdc38be6ebf49322","first_computed_at":"2026-05-20T00:03:13.716203Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:03:13.716203Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"ykMPkiiLVH260eWwmRUZxE5tZz+kqK7phzr5wy5nHqkaFaGRUlhy5Lt3q5d9/LA8eI0lnmC9Idw67qKaMwrQCw==","signature_status":"signed_v1","signed_at":"2026-05-20T00:03:13.717281Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.02167","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:835ecadef92dd3b028a3ca7290f3fa83d9b212a41ac8216fde577cffe234bf81","sha256:94678ad2141f371ebea6db9c3f60fc69ea2ffb351dbfc2779072aa3b30e6d33a","sha256:13d4eb732766259573c9813e1b36b5f42f1c366568e5fb3e3990a80baaaea866"],"state_sha256":"0d504d8a54bc9d9caff68a73e4b8c5fdfaed2794b2712ab01ebb7796de80f702"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/O3y36Nq+JbKMjn9MpeerdMam8v7CUhac3qXomqtc9uOAfyUwXyOAJgIB1SPto7CCM0FxZ2tXLcVNArh+Fb0Bw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-21T16:08:11.289862Z","bundle_sha256":"178df08545454c5aab07950752414263b7e9fc81f607b6b4513c8b6fba8675f9"}}