{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:HV4NPTAHZKI5WYKFUZSSEZHAUM","short_pith_number":"pith:HV4NPTAH","canonical_record":{"source":{"id":"2605.14113","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T20:57:37Z","cross_cats_sorted":["cs.AI","cs.LG","cs.MA"],"title_canon_sha256":"d4c4c67bbcd1800c7471a71852dd29ff768bf2d6a9c353513b613fa253dfa05d","abstract_canon_sha256":"742ea7beee70ea94b2a6818ef7a2dd4e26a53cc52c26ecb21b097f1f5e534998"},"schema_version":"1.0"},"canonical_sha256":"3d78d7cc07ca91db6145a6652264e0a3379e4f33d7548df74804c252dbabc3d4","source":{"kind":"arxiv","id":"2605.14113","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14113","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14113v1","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14113","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"pith_short_12","alias_value":"HV4NPTAHZKI5","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HV4NPTAHZKI5WYKF","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HV4NPTAH","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:HV4NPTAHZKI5WYKFUZSSEZHAUM","target":"record","payload":{"canonical_record":{"source":{"id":"2605.14113","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T20:57:37Z","cross_cats_sorted":["cs.AI","cs.LG","cs.MA"],"title_canon_sha256":"d4c4c67bbcd1800c7471a71852dd29ff768bf2d6a9c353513b613fa253dfa05d","abstract_canon_sha256":"742ea7beee70ea94b2a6818ef7a2dd4e26a53cc52c26ecb21b097f1f5e534998"},"schema_version":"1.0"},"canonical_sha256":"3d78d7cc07ca91db6145a6652264e0a3379e4f33d7548df74804c252dbabc3d4","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:11.978911Z","signature_b64":"+/vw7o4lPl6yWWCO1Z74/83zJdf+MJdGFtxE1mIDnAoduZevdhc9yB9etVSE6+8900k4TdZhKgXJwa7hjzXeBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3d78d7cc07ca91db6145a6652264e0a3379e4f33d7548df74804c252dbabc3d4","last_reissued_at":"2026-05-17T23:39:11.978435Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:11.978435Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.14113","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":"2026-05-17T23:39:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ry3jZ9Ix98Heb4O8a5J+IAE/JAy7s+lCDFSYoIw5Yrvs6NivN1SAqJ2SKUVBF9GLvkwQ8GR86rVE5bn0ft/tCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-28T02:46:12.252408Z"},"content_sha256":"a0601032af030d3d048ebfec2cf518d1efc4ecaa3ab9a5d8b127aae99362e0c6","schema_version":"1.0","event_id":"sha256:a0601032af030d3d048ebfec2cf518d1efc4ecaa3ab9a5d8b127aae99362e0c6"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:HV4NPTAHZKI5WYKFUZSSEZHAUM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"ProtoMedAgent: Multimodal Clinical Interpretability via Privacy-Aware Agentic Workflows","license":"http://creativecommons.org/licenses/by/4.0/","headline":"ProtoMedAgent constrains LLM clinical reports to a neuro-symbolic bottleneck to reach 91.2% faithfulness while cutting membership inference risks.","cross_cats":["cs.AI","cs.LG","cs.MA"],"primary_cat":"cs.CV","authors_text":"Alvaro Lopez Pellicer, Eduardo Soares, Jemma Kerns, Marwan Bukhari, Plamen Angelov, Yi Li","submitted_at":"2026-05-13T20:57:37Z","abstract_excerpt":"While interpretable prototype networks offer compelling case-based reasoning for clinical diagnostics, their raw continuous outputs lack the semantic structure required for medical documentation. Bridging this gap via standard Retrieval-Augmented Generation (RAG) routinely triggers ``retrieval sycophancy,'' where Large Language Models (LLMs) hallucinate post-hoc rationalizations to align with visual predictions. We introduce ProtoMedAgent, a framework that formalizes multimodal clinical reporting as an iterative, zero-gradient test-time optimization problem over a strict neuro-symbolic bottlen"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"ProtoMedAgent achieves 91.2% Comparison Set Faithfulness where it fundamentally outperforms standard RAG (46.2%). It leverages a binding ℓ-diversity phase transition to reduce artifact-level membership inference risks by an absolute 9.8%.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the iterative zero-gradient test-time optimization over the neuro-symbolic bottleneck with the Scribe-Critic loop can mathematically preclude unsupported narrative claims in all cases, and that the semantic privacy gate effectively bounds disclosure without compromising utility.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ProtoMedAgent uses a privacy-aware agentic workflow with neuro-symbolic bottlenecks to achieve 91.2% faithfulness in clinical report generation, significantly outperforming standard RAG methods on a large patient cohort.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"ProtoMedAgent constrains LLM clinical reports to a neuro-symbolic bottleneck to reach 91.2% faithfulness while cutting membership inference risks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ac8e900b9d1db8c7acf799dd8c8558de4b9622ed9f66d3f8d95fb45991d1ee89"},"source":{"id":"2605.14113","kind":"arxiv","version":1},"verdict":{"id":"a79e452c-e36e-44a2-ac7a-a887e84bb77e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T05:05:37.166680Z","strongest_claim":"ProtoMedAgent achieves 91.2% Comparison Set Faithfulness where it fundamentally outperforms standard RAG (46.2%). It leverages a binding ℓ-diversity phase transition to reduce artifact-level membership inference risks by an absolute 9.8%.","one_line_summary":"ProtoMedAgent uses a privacy-aware agentic workflow with neuro-symbolic bottlenecks to achieve 91.2% faithfulness in clinical report generation, significantly outperforming standard RAG methods on a large patient cohort.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the iterative zero-gradient test-time optimization over the neuro-symbolic bottleneck with the Scribe-Critic loop can mathematically preclude unsupported narrative claims in all cases, and that the semantic privacy gate effectively bounds disclosure without compromising utility.","pith_extraction_headline":"ProtoMedAgent constrains LLM clinical reports to a neuro-symbolic bottleneck to reach 91.2% faithfulness while cutting membership inference risks."},"references":{"count":45,"sample":[{"doi":"","year":null,"title":"Towards explainable deep neural networks (xdnn).Neural Networks, 130:185–194,","work_id":"d70e73e4-4c2f-48a1-aa3e-e1a52e527da9","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2026,"title":"Trace: Tem- poral radiology with anatomical change explanation for grounded x-ray report generation, 2026","work_id":"3bd23b27-6cbe-4717-84f2-a65f89cd8ed4","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Large language models in radiology reporting-a sys- tematic review of performance, limitations, and clinical im- plications.Intelligence-Based Medicine, page 100287, 2025","work_id":"5433112d-e498-46eb-9a0c-a87c4dacee92","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2019,"title":"This looks like that: Deep learning for interpretable image recognition","work_id":"ca5930ae-0eb6-4e23-b42a-8aa36f852774","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Meditron-70b: Scaling medical pretraining for large language models, 2023","work_id":"b2104b98-71db-4fcb-badd-e7ba0b891715","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":45,"snapshot_sha256":"a57d650f435c4580493373b802602e4bb8dbc52cf5b3d757ab2f9bb6f42edb0d","internal_anchors":4},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1034ddad0351ef81e02b0fab7ad47ee6ddde7ba43558e0ed651273be9e4414bd"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"a79e452c-e36e-44a2-ac7a-a887e84bb77e"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:11Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"WF3zWJk1X4CWHeFk1XcZS9y54fFuzlwMIa3LmM6cvHo2mjKVdBzz2DiIeMJ83A+KMM/EUM10PADSEJsaK8rmDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-28T02:46:12.253067Z"},"content_sha256":"37a97c32c936c60de98af79ea5d1790ddd70c3f5c88a068c5e1023e86f92bb74","schema_version":"1.0","event_id":"sha256:37a97c32c936c60de98af79ea5d1790ddd70c3f5c88a068c5e1023e86f92bb74"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM/bundle.json","state_url":"https://pith.science/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM/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":"2026-05-28T02:46:12Z","links":{"resolver":"https://pith.science/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM","bundle":"https://pith.science/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM/bundle.json","state":"https://pith.science/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/HV4NPTAHZKI5WYKFUZSSEZHAUM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:HV4NPTAHZKI5WYKFUZSSEZHAUM","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":"742ea7beee70ea94b2a6818ef7a2dd4e26a53cc52c26ecb21b097f1f5e534998","cross_cats_sorted":["cs.AI","cs.LG","cs.MA"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T20:57:37Z","title_canon_sha256":"d4c4c67bbcd1800c7471a71852dd29ff768bf2d6a9c353513b613fa253dfa05d"},"schema_version":"1.0","source":{"id":"2605.14113","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14113","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14113v1","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14113","created_at":"2026-05-17T23:39:11Z"},{"alias_kind":"pith_short_12","alias_value":"HV4NPTAHZKI5","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"HV4NPTAHZKI5WYKF","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"HV4NPTAH","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:37a97c32c936c60de98af79ea5d1790ddd70c3f5c88a068c5e1023e86f92bb74","target":"graph","created_at":"2026-05-17T23:39:11Z","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":"ProtoMedAgent achieves 91.2% Comparison Set Faithfulness where it fundamentally outperforms standard RAG (46.2%). It leverages a binding ℓ-diversity phase transition to reduce artifact-level membership inference risks by an absolute 9.8%."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the iterative zero-gradient test-time optimization over the neuro-symbolic bottleneck with the Scribe-Critic loop can mathematically preclude unsupported narrative claims in all cases, and that the semantic privacy gate effectively bounds disclosure without compromising utility."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"ProtoMedAgent uses a privacy-aware agentic workflow with neuro-symbolic bottlenecks to achieve 91.2% faithfulness in clinical report generation, significantly outperforming standard RAG methods on a large patient cohort."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"ProtoMedAgent constrains LLM clinical reports to a neuro-symbolic bottleneck to reach 91.2% faithfulness while cutting membership inference risks."}],"snapshot_sha256":"ac8e900b9d1db8c7acf799dd8c8558de4b9622ed9f66d3f8d95fb45991d1ee89"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"1034ddad0351ef81e02b0fab7ad47ee6ddde7ba43558e0ed651273be9e4414bd"},"paper":{"abstract_excerpt":"While interpretable prototype networks offer compelling case-based reasoning for clinical diagnostics, their raw continuous outputs lack the semantic structure required for medical documentation. Bridging this gap via standard Retrieval-Augmented Generation (RAG) routinely triggers ``retrieval sycophancy,'' where Large Language Models (LLMs) hallucinate post-hoc rationalizations to align with visual predictions. We introduce ProtoMedAgent, a framework that formalizes multimodal clinical reporting as an iterative, zero-gradient test-time optimization problem over a strict neuro-symbolic bottlen","authors_text":"Alvaro Lopez Pellicer, Eduardo Soares, Jemma Kerns, Marwan Bukhari, Plamen Angelov, Yi Li","cross_cats":["cs.AI","cs.LG","cs.MA"],"headline":"ProtoMedAgent constrains LLM clinical reports to a neuro-symbolic bottleneck to reach 91.2% faithfulness while cutting membership inference risks.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T20:57:37Z","title":"ProtoMedAgent: Multimodal Clinical Interpretability via Privacy-Aware Agentic Workflows"},"references":{"count":45,"internal_anchors":4,"resolved_work":45,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Towards explainable deep neural networks (xdnn).Neural Networks, 130:185–194,","work_id":"d70e73e4-4c2f-48a1-aa3e-e1a52e527da9","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Trace: Tem- poral radiology with anatomical change explanation for grounded x-ray report generation, 2026","work_id":"3bd23b27-6cbe-4717-84f2-a65f89cd8ed4","year":2026},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Large language models in radiology reporting-a sys- tematic review of performance, limitations, and clinical im- plications.Intelligence-Based Medicine, page 100287, 2025","work_id":"5433112d-e498-46eb-9a0c-a87c4dacee92","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"This looks like that: Deep learning for interpretable image recognition","work_id":"ca5930ae-0eb6-4e23-b42a-8aa36f852774","year":2019},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Meditron-70b: Scaling medical pretraining for large language models, 2023","work_id":"b2104b98-71db-4fcb-badd-e7ba0b891715","year":2023}],"snapshot_sha256":"a57d650f435c4580493373b802602e4bb8dbc52cf5b3d757ab2f9bb6f42edb0d"},"source":{"id":"2605.14113","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T05:05:37.166680Z","id":"a79e452c-e36e-44a2-ac7a-a887e84bb77e","model_set":{"reader":"grok-4.3"},"one_line_summary":"ProtoMedAgent uses a privacy-aware agentic workflow with neuro-symbolic bottlenecks to achieve 91.2% faithfulness in clinical report generation, significantly outperforming standard RAG methods on a large patient cohort.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"ProtoMedAgent constrains LLM clinical reports to a neuro-symbolic bottleneck to reach 91.2% faithfulness while cutting membership inference risks.","strongest_claim":"ProtoMedAgent achieves 91.2% Comparison Set Faithfulness where it fundamentally outperforms standard RAG (46.2%). It leverages a binding ℓ-diversity phase transition to reduce artifact-level membership inference risks by an absolute 9.8%.","weakest_assumption":"That the iterative zero-gradient test-time optimization over the neuro-symbolic bottleneck with the Scribe-Critic loop can mathematically preclude unsupported narrative claims in all cases, and that the semantic privacy gate effectively bounds disclosure without compromising utility."}},"verdict_id":"a79e452c-e36e-44a2-ac7a-a887e84bb77e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:a0601032af030d3d048ebfec2cf518d1efc4ecaa3ab9a5d8b127aae99362e0c6","target":"record","created_at":"2026-05-17T23:39:11Z","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":"742ea7beee70ea94b2a6818ef7a2dd4e26a53cc52c26ecb21b097f1f5e534998","cross_cats_sorted":["cs.AI","cs.LG","cs.MA"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-13T20:57:37Z","title_canon_sha256":"d4c4c67bbcd1800c7471a71852dd29ff768bf2d6a9c353513b613fa253dfa05d"},"schema_version":"1.0","source":{"id":"2605.14113","kind":"arxiv","version":1}},"canonical_sha256":"3d78d7cc07ca91db6145a6652264e0a3379e4f33d7548df74804c252dbabc3d4","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3d78d7cc07ca91db6145a6652264e0a3379e4f33d7548df74804c252dbabc3d4","first_computed_at":"2026-05-17T23:39:11.978435Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:11.978435Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"+/vw7o4lPl6yWWCO1Z74/83zJdf+MJdGFtxE1mIDnAoduZevdhc9yB9etVSE6+8900k4TdZhKgXJwa7hjzXeBQ==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:11.978911Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14113","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a0601032af030d3d048ebfec2cf518d1efc4ecaa3ab9a5d8b127aae99362e0c6","sha256:37a97c32c936c60de98af79ea5d1790ddd70c3f5c88a068c5e1023e86f92bb74"],"state_sha256":"79e519dc1ae634b88697204b60f1fda7eed98f93b18c231d6c60a129fd162dea"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"5pPfvuydCUj1LqLjutwgGumln/O6+S6DlaUNVkTSqmh+bHeDGpKdWJaYvrIW2/NO4TQ/UizaJWFwd4AJ3TLJDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-28T02:46:12.255662Z","bundle_sha256":"4f1e0c7626e41ee7f635e2c9904a74de5a6ece8a088542d4a816658428a44ee5"}}