{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:RA2L3U7SF5FSQ4DDFLXTXW6AQG","short_pith_number":"pith:RA2L3U7S","canonical_record":{"source":{"id":"2605.15306","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:19:40Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"3f60ba417194fd8793176695cc9493b0b65619759fd7e0d4a7a5c08afa5e4ff9","abstract_canon_sha256":"6fb228d36b303dc0fc7e159fa17f9bfe859ca9a15431dea37dcd728eb523e3a2"},"schema_version":"1.0"},"canonical_sha256":"8834bdd3f22f4b2870632aef3bdbc081abecf49f8714f57920a27b38a7e95aa3","source":{"kind":"arxiv","id":"2605.15306","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.15306","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"arxiv_version","alias_value":"2605.15306v1","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15306","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"pith_short_12","alias_value":"RA2L3U7SF5FS","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"pith_short_16","alias_value":"RA2L3U7SF5FSQ4DD","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"pith_short_8","alias_value":"RA2L3U7S","created_at":"2026-05-20T00:00:51Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:RA2L3U7SF5FSQ4DDFLXTXW6AQG","target":"record","payload":{"canonical_record":{"source":{"id":"2605.15306","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:19:40Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"3f60ba417194fd8793176695cc9493b0b65619759fd7e0d4a7a5c08afa5e4ff9","abstract_canon_sha256":"6fb228d36b303dc0fc7e159fa17f9bfe859ca9a15431dea37dcd728eb523e3a2"},"schema_version":"1.0"},"canonical_sha256":"8834bdd3f22f4b2870632aef3bdbc081abecf49f8714f57920a27b38a7e95aa3","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:00:51.756209Z","signature_b64":"JIXL+E6XsjG/OgnNVlXGb6C4/HuaV1hPA7Hbam1IO39iKWCL6plJuod63WH3+ECfHVbksy2qx8sjicZR5SCEDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8834bdd3f22f4b2870632aef3bdbc081abecf49f8714f57920a27b38a7e95aa3","last_reissued_at":"2026-05-20T00:00:51.755550Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:00:51.755550Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.15306","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":"2026-05-20T00:00:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jhZ/kGBDUONCSLRBJI0HfPmIRVhIe328dWqveHLxm7+SfkOoWHfB97wLL2c94h58nIYcGNDFPnDCd31wbWZlBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-22T19:06:49.087034Z"},"content_sha256":"0d814016f1e00414f1c8cd07a2becacaaed64ad8819c871c29cd36e3b37dac9f","schema_version":"1.0","event_id":"sha256:0d814016f1e00414f1c8cd07a2becacaaed64ad8819c871c29cd36e3b37dac9f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:RA2L3U7SF5FSQ4DDFLXTXW6AQG","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"How Data Augmentation Shapes Neural Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Data augmentation steers neural representations along distinct, predictable trajectories in an invariant shape space.","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Alex H. Williams, Sarah E. Harvey, Tianxiao He","submitted_at":"2026-05-14T18:19:40Z","abstract_excerpt":"Data augmentation is widely recognized for improving generalization in deep networks, yet its impact on the geometry of learned representations remains poorly understood. In this work, we characterize how different data augmentation strategies reshape internal representations in neural networks. Using tools from shape analysis, we embed network hidden representations into a metric space where distance is invariant to scaling, translation, rotation and reflection. We show that increasing augmentation strength leads to well-behaved trajectories in this space, and that different augmentation type"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Increasing augmentation strength leads to well-behaved trajectories in this space, and that different augmentation types steer representations in distinct directions. Moreover, insights from neural geometry can predict which representations provide the most improvement when ensembling models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the chosen invariant metric embedding of hidden representations captures the geometric properties most relevant to generalization and ensembling performance.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Data augmentation produces well-behaved trajectories in shape-invariant representation space, with augmentation type steering distinct directions and geometry predicting ensembling gains.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Data augmentation steers neural representations along distinct, predictable trajectories in an invariant shape space.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"034ab52973867fcc5a22b0f88a783a835f01556d9689dea739567a70a3076bf6"},"source":{"id":"2605.15306","kind":"arxiv","version":1},"verdict":{"id":"eeeb3503-8698-4aa6-8cf9-4f43319e8d50","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T16:20:15.777938Z","strongest_claim":"Increasing augmentation strength leads to well-behaved trajectories in this space, and that different augmentation types steer representations in distinct directions. Moreover, insights from neural geometry can predict which representations provide the most improvement when ensembling models.","one_line_summary":"Data augmentation produces well-behaved trajectories in shape-invariant representation space, with augmentation type steering distinct directions and geometry predicting ensembling gains.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the chosen invariant metric embedding of hidden representations captures the geometric properties most relevant to generalization and ensembling performance.","pith_extraction_headline":"Data augmentation steers neural representations along distinct, predictable trajectories in an invariant shape space."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15306/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:31:18.345789Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T16:27:03.339934Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.224713Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.777400Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"4831b78ca5ee7803902edce8f636113dab0d53fcf0f31ac477558f87c162d7c2"},"references":{"count":29,"sample":[{"doi":"","year":2018,"title":"month = apr, year =","work_id":"004bf78b-4faf-4d90-961e-973b64c4cc53","ref_index":1,"cited_arxiv_id":"1803.09820","is_internal_anchor":true},{"doi":"10.1186/s40537-019-0197-0","year":2019,"title":"Khoshgoftaar, A survey on Image Data Augmentation for Deep Learning, J Big Data, 6 (2019), pp","work_id":"f2e71cc4-6f55-4bdc-af93-530077477746","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"A Kernel Theory of Modern Data Augmentation","work_id":"e96c1724-7f8c-4204-9aa0-7fb964b5eddf","ref_index":3,"cited_arxiv_id":"1803.06084","is_internal_anchor":true},{"doi":"","year":2015,"title":"Understanding Image Representations by Measuring Their Equivariance and Equivalence","work_id":"dcf667af-ef54-4b55-93bf-d03c9a174971","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"A group-theoretic framework for data augmentation","work_id":"301ced0e-2dc9-4844-8b39-2204b0efe1ed","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":29,"snapshot_sha256":"d07231aa1c5461fe8ddf7bddf60cc8c569dd805292f7f1cd7024d30dc7138a6e","internal_anchors":6},"formal_canon":{"evidence_count":2,"snapshot_sha256":"bbdd495f19a7b469aa7ee8feb6940399df97a9368bf87d233b0459bee40e36fd"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"eeeb3503-8698-4aa6-8cf9-4f43319e8d50"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:00:51Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6lh0zhpa9jqbI/riVzDcwnkPppF9Y3IuvrXlh+/34J53S53g8eeBFAagIbI1D7Y44mzMgvGpoWOwXHpDnw+DDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-22T19:06:49.087695Z"},"content_sha256":"476ed14f3d3fdaf348856298bf6cbf3afb7d79424b08157a33892746dde65e33","schema_version":"1.0","event_id":"sha256:476ed14f3d3fdaf348856298bf6cbf3afb7d79424b08157a33892746dde65e33"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG/bundle.json","state_url":"https://pith.science/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG/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":"2026-05-22T19:06:49Z","links":{"resolver":"https://pith.science/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG","bundle":"https://pith.science/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG/bundle.json","state":"https://pith.science/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RA2L3U7SF5FSQ4DDFLXTXW6AQG/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:RA2L3U7SF5FSQ4DDFLXTXW6AQG","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":"6fb228d36b303dc0fc7e159fa17f9bfe859ca9a15431dea37dcd728eb523e3a2","cross_cats_sorted":["stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:19:40Z","title_canon_sha256":"3f60ba417194fd8793176695cc9493b0b65619759fd7e0d4a7a5c08afa5e4ff9"},"schema_version":"1.0","source":{"id":"2605.15306","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.15306","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"arxiv_version","alias_value":"2605.15306v1","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15306","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"pith_short_12","alias_value":"RA2L3U7SF5FS","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"pith_short_16","alias_value":"RA2L3U7SF5FSQ4DD","created_at":"2026-05-20T00:00:51Z"},{"alias_kind":"pith_short_8","alias_value":"RA2L3U7S","created_at":"2026-05-20T00:00:51Z"}],"graph_snapshots":[{"event_id":"sha256:476ed14f3d3fdaf348856298bf6cbf3afb7d79424b08157a33892746dde65e33","target":"graph","created_at":"2026-05-20T00:00:51Z","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":"Increasing augmentation strength leads to well-behaved trajectories in this space, and that different augmentation types steer representations in distinct directions. Moreover, insights from neural geometry can predict which representations provide the most improvement when ensembling models."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the chosen invariant metric embedding of hidden representations captures the geometric properties most relevant to generalization and ensembling performance."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Data augmentation produces well-behaved trajectories in shape-invariant representation space, with augmentation type steering distinct directions and geometry predicting ensembling gains."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Data augmentation steers neural representations along distinct, predictable trajectories in an invariant shape space."}],"snapshot_sha256":"034ab52973867fcc5a22b0f88a783a835f01556d9689dea739567a70a3076bf6"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"bbdd495f19a7b469aa7ee8feb6940399df97a9368bf87d233b0459bee40e36fd"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T16:31:18.345789Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T16:27:03.339934Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T14:41:54.224713Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.777400Z","status":"skipped","version":"1.0.0"}],"endpoint":"/pith/2605.15306/integrity.json","findings":[],"snapshot_sha256":"4831b78ca5ee7803902edce8f636113dab0d53fcf0f31ac477558f87c162d7c2","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Data augmentation is widely recognized for improving generalization in deep networks, yet its impact on the geometry of learned representations remains poorly understood. In this work, we characterize how different data augmentation strategies reshape internal representations in neural networks. Using tools from shape analysis, we embed network hidden representations into a metric space where distance is invariant to scaling, translation, rotation and reflection. We show that increasing augmentation strength leads to well-behaved trajectories in this space, and that different augmentation type","authors_text":"Alex H. Williams, Sarah E. Harvey, Tianxiao He","cross_cats":["stat.ML"],"headline":"Data augmentation steers neural representations along distinct, predictable trajectories in an invariant shape space.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:19:40Z","title":"How Data Augmentation Shapes Neural Representations"},"references":{"count":29,"internal_anchors":6,"resolved_work":29,"sample":[{"cited_arxiv_id":"1803.09820","doi":"","is_internal_anchor":true,"ref_index":1,"title":"month = apr, year =","work_id":"004bf78b-4faf-4d90-961e-973b64c4cc53","year":2018},{"cited_arxiv_id":"","doi":"10.1186/s40537-019-0197-0","is_internal_anchor":false,"ref_index":2,"title":"Khoshgoftaar, A survey on Image Data Augmentation for Deep Learning, J Big Data, 6 (2019), pp","work_id":"f2e71cc4-6f55-4bdc-af93-530077477746","year":2019},{"cited_arxiv_id":"1803.06084","doi":"","is_internal_anchor":true,"ref_index":3,"title":"A Kernel Theory of Modern Data Augmentation","work_id":"e96c1724-7f8c-4204-9aa0-7fb964b5eddf","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Understanding Image Representations by Measuring Their Equivariance and Equivalence","work_id":"dcf667af-ef54-4b55-93bf-d03c9a174971","year":2015},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"A group-theoretic framework for data augmentation","work_id":"301ced0e-2dc9-4844-8b39-2204b0efe1ed","year":2020}],"snapshot_sha256":"d07231aa1c5461fe8ddf7bddf60cc8c569dd805292f7f1cd7024d30dc7138a6e"},"source":{"id":"2605.15306","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T16:20:15.777938Z","id":"eeeb3503-8698-4aa6-8cf9-4f43319e8d50","model_set":{"reader":"grok-4.3"},"one_line_summary":"Data augmentation produces well-behaved trajectories in shape-invariant representation space, with augmentation type steering distinct directions and geometry predicting ensembling gains.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Data augmentation steers neural representations along distinct, predictable trajectories in an invariant shape space.","strongest_claim":"Increasing augmentation strength leads to well-behaved trajectories in this space, and that different augmentation types steer representations in distinct directions. Moreover, insights from neural geometry can predict which representations provide the most improvement when ensembling models.","weakest_assumption":"That the chosen invariant metric embedding of hidden representations captures the geometric properties most relevant to generalization and ensembling performance."}},"verdict_id":"eeeb3503-8698-4aa6-8cf9-4f43319e8d50"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:0d814016f1e00414f1c8cd07a2becacaaed64ad8819c871c29cd36e3b37dac9f","target":"record","created_at":"2026-05-20T00:00:51Z","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":"6fb228d36b303dc0fc7e159fa17f9bfe859ca9a15431dea37dcd728eb523e3a2","cross_cats_sorted":["stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T18:19:40Z","title_canon_sha256":"3f60ba417194fd8793176695cc9493b0b65619759fd7e0d4a7a5c08afa5e4ff9"},"schema_version":"1.0","source":{"id":"2605.15306","kind":"arxiv","version":1}},"canonical_sha256":"8834bdd3f22f4b2870632aef3bdbc081abecf49f8714f57920a27b38a7e95aa3","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8834bdd3f22f4b2870632aef3bdbc081abecf49f8714f57920a27b38a7e95aa3","first_computed_at":"2026-05-20T00:00:51.755550Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:00:51.755550Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"JIXL+E6XsjG/OgnNVlXGb6C4/HuaV1hPA7Hbam1IO39iKWCL6plJuod63WH3+ECfHVbksy2qx8sjicZR5SCEDw==","signature_status":"signed_v1","signed_at":"2026-05-20T00:00:51.756209Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.15306","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:0d814016f1e00414f1c8cd07a2becacaaed64ad8819c871c29cd36e3b37dac9f","sha256:476ed14f3d3fdaf348856298bf6cbf3afb7d79424b08157a33892746dde65e33"],"state_sha256":"885988da120d14b4c915078895739b5a6ff520b8858d1213ffc3b266525aff1d"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8DA7BTGawJgjAmjYeogTH2Nr5Uj/FzGFgAZkr7bOTpZP3awtNWhF1+Twd6IkfgpUnsBhIVeRpSKeXXyGSQf4Bg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-22T19:06:49.090622Z","bundle_sha256":"5cbe7150c18d8b9eee126dfd5039df3acd27b6ea8f74eeef9ff603fbbf2dd997"}}