{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2020:2PKZKQ3RNR2EBKZ6UWAEXCZQSR","short_pith_number":"pith:2PKZKQ3R","canonical_record":{"source":{"id":"2010.14701","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-28T02:17:24Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"1f943d19ab5e7a38eac39acd563156ca4196d2797172c4b2ac8a298f40aa7643","abstract_canon_sha256":"50db097e855a42be45810223f70bda72257bbf6cdf78c3452587fe3b8c741924"},"schema_version":"1.0"},"canonical_sha256":"d3d59543716c7440ab3ea5804b8b309450156f477e95f20f00ade73f9430539c","source":{"kind":"arxiv","id":"2010.14701","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2010.14701","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"arxiv_version","alias_value":"2010.14701v2","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.14701","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"pith_short_12","alias_value":"2PKZKQ3RNR2E","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"pith_short_16","alias_value":"2PKZKQ3RNR2EBKZ6","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"pith_short_8","alias_value":"2PKZKQ3R","created_at":"2026-07-05T01:49:38Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2020:2PKZKQ3RNR2EBKZ6UWAEXCZQSR","target":"record","payload":{"canonical_record":{"source":{"id":"2010.14701","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-28T02:17:24Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"1f943d19ab5e7a38eac39acd563156ca4196d2797172c4b2ac8a298f40aa7643","abstract_canon_sha256":"50db097e855a42be45810223f70bda72257bbf6cdf78c3452587fe3b8c741924"},"schema_version":"1.0"},"canonical_sha256":"d3d59543716c7440ab3ea5804b8b309450156f477e95f20f00ade73f9430539c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:49:38.993111Z","signature_b64":"yW1RFQKqKanIyapciTUO1sGh6fg76OGcnSTZqSgJ61WNkT6IdI/4w3Ja5uIQwTa7e2H06vmSz/dc9o8rM5jQDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d3d59543716c7440ab3ea5804b8b309450156f477e95f20f00ade73f9430539c","last_reissued_at":"2026-07-05T01:49:38.992685Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:49:38.992685Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2010.14701","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":"2026-07-05T01:49:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"rd6L5BCky8NEg4I7Z1SVN5/WdaDNY3/enshlbaTXch5YvEof5b1Gv3zSWD64VDrig/rIECIrKCYqa3bMu2LQBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T13:37:15.640548Z"},"content_sha256":"e3a6b78fa7e20ef099836eac0605fcea0fe92a578e5d31a1e7031310ea844122","schema_version":"1.0","event_id":"sha256:e3a6b78fa7e20ef099836eac0605fcea0fe92a578e5d31a1e7031310ea844122"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2020:2PKZKQ3RNR2EBKZ6UWAEXCZQSR","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Scaling Laws for Autoregressive Generative Modeling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Autoregressive Transformers improve via power-law-plus-constant scaling laws for cross-entropy loss across image, video, text-image, and math domains, with optimal size depending on compute through nearly universal exponents.","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Aditya Ramesh, Alec Radford, Benjamin Mann, Chris Hallacy, Christopher Hesse, Daniel M. Ziegler, Dario Amodei, Heewoo Jun, Jacob Jackson, Jared Kaplan, John Schulman, Mark Chen, Mor Katz, Nick Ryder, Prafulla Dhariwal, Sam McCandlish, Scott Gray, Tom B. Brown, Tom Henighan","submitted_at":"2020-10-28T02:17:24Z","abstract_excerpt":"We identify empirical scaling laws for the cross-entropy loss in four domains: generative image modeling, video modeling, multimodal image$\\leftrightarrow$text models, and mathematical problem solving. In all cases autoregressive Transformers smoothly improve in performance as model size and compute budgets increase, following a power-law plus constant scaling law. The optimal model size also depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains.\n  The cross-entropy loss has an information theoretic interpretation as $S($True$) + D_"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The optimal model size depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the power-law-plus-constant functional form continues to hold outside the range of model sizes and compute budgets actually measured in the experiments.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Autoregressive transformers follow power-law scaling laws for cross-entropy loss with nearly universal exponents relating optimal model size to compute budget across four domains.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Autoregressive Transformers improve via power-law-plus-constant scaling laws for cross-entropy loss across image, video, text-image, and math domains, with optimal size depending on compute through nearly universal exponents.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a81a7db543c74d84e63ecaa8220c1f23668451cfb6b4457d6b737d8ada75c663"},"source":{"id":"2010.14701","kind":"arxiv","version":2},"verdict":{"id":"c673df1b-51cd-40ed-b1f0-dbcd8b110e14","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T07:44:52.955019Z","strongest_claim":"The optimal model size depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains.","one_line_summary":"Autoregressive transformers follow power-law scaling laws for cross-entropy loss with nearly universal exponents relating optimal model size to compute budget across four domains.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the power-law-plus-constant functional form continues to hold outside the range of model sizes and compute budgets actually measured in the experiments.","pith_extraction_headline":"Autoregressive Transformers improve via power-law-plus-constant scaling laws for cross-entropy loss across image, video, text-image, and math domains, with optimal size depending on compute through nearly universal exponents."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2010.14701/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":29,"sample":[{"doi":"","year":2020,"title":"Language Models are Few-Shot Learners","work_id":"214732c0-2edd-44a0-af9e-28184a2b8279","ref_index":1,"cited_arxiv_id":"2005.14165","is_internal_anchor":true},{"doi":"","year":1904,"title":"Generating Long Sequences with Sparse Transformers","work_id":"c5b81688-45ee-4a9a-b095-e6290f45cb6c","ref_index":2,"cited_arxiv_id":"1904.10509","is_internal_anchor":true},{"doi":"","year":2017,"title":"A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets","work_id":"b28da76e-30a0-4103-a956-55ca47f3e243","ref_index":3,"cited_arxiv_id":"1707.08819","is_internal_anchor":false},{"doi":"","year":2020,"title":"Generative pretraining from pixels","work_id":"e47419f8-22d5-401d-899d-bcc702f9885f","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Jukebox: A Generative Model for Music","work_id":"aeb5293c-aee5-4e2d-ae57-668eb930893e","ref_index":5,"cited_arxiv_id":"2005.00341","is_internal_anchor":false}],"resolved_work":29,"snapshot_sha256":"9d9f4af64c1b8a1da26c8877cbed0de0ddc8513600fc74125780a7db5ff4228d","internal_anchors":5},"formal_canon":{"evidence_count":2,"snapshot_sha256":"789672bef06716a6b7091e5d27b284bedc746e10c3a209565b85d9f64b09cb68"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"c673df1b-51cd-40ed-b1f0-dbcd8b110e14"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T01:49:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+g6X+xQBPC8NrPRBLI0lxl0hSXDmYH0g7zgrEJE7NcB55Zz/0fhfWa4qofw+O7iASrGtgh0vMCIa1+fFFzUBCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T13:37:15.641379Z"},"content_sha256":"7d4cc30999b5f93b44a42ffbce876880697c6723d398d6b3496e478e12c6884a","schema_version":"1.0","event_id":"sha256:7d4cc30999b5f93b44a42ffbce876880697c6723d398d6b3496e478e12c6884a"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR/bundle.json","state_url":"https://pith.science/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR/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":"2026-08-04T13:37:15Z","links":{"resolver":"https://pith.science/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR","bundle":"https://pith.science/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR/bundle.json","state":"https://pith.science/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR/state.json","well_known_bundle":"https://pith.science/.well-known/pith/2PKZKQ3RNR2EBKZ6UWAEXCZQSR/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2020:2PKZKQ3RNR2EBKZ6UWAEXCZQSR","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":"50db097e855a42be45810223f70bda72257bbf6cdf78c3452587fe3b8c741924","cross_cats_sorted":["cs.CL","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-28T02:17:24Z","title_canon_sha256":"1f943d19ab5e7a38eac39acd563156ca4196d2797172c4b2ac8a298f40aa7643"},"schema_version":"1.0","source":{"id":"2010.14701","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2010.14701","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"arxiv_version","alias_value":"2010.14701v2","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.14701","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"pith_short_12","alias_value":"2PKZKQ3RNR2E","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"pith_short_16","alias_value":"2PKZKQ3RNR2EBKZ6","created_at":"2026-07-05T01:49:38Z"},{"alias_kind":"pith_short_8","alias_value":"2PKZKQ3R","created_at":"2026-07-05T01:49:38Z"}],"graph_snapshots":[{"event_id":"sha256:7d4cc30999b5f93b44a42ffbce876880697c6723d398d6b3496e478e12c6884a","target":"graph","created_at":"2026-07-05T01:49:38Z","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 optimal model size depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the power-law-plus-constant functional form continues to hold outside the range of model sizes and compute budgets actually measured in the experiments."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Autoregressive transformers follow power-law scaling laws for cross-entropy loss with nearly universal exponents relating optimal model size to compute budget across four domains."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Autoregressive Transformers improve via power-law-plus-constant scaling laws for cross-entropy loss across image, video, text-image, and math domains, with optimal size depending on compute through nearly universal exponents."}],"snapshot_sha256":"a81a7db543c74d84e63ecaa8220c1f23668451cfb6b4457d6b737d8ada75c663"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"789672bef06716a6b7091e5d27b284bedc746e10c3a209565b85d9f64b09cb68"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2010.14701/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We identify empirical scaling laws for the cross-entropy loss in four domains: generative image modeling, video modeling, multimodal image$\\leftrightarrow$text models, and mathematical problem solving. In all cases autoregressive Transformers smoothly improve in performance as model size and compute budgets increase, following a power-law plus constant scaling law. The optimal model size also depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains.\n  The cross-entropy loss has an information theoretic interpretation as $S($True$) + D_","authors_text":"Aditya Ramesh, Alec Radford, Benjamin Mann, Chris Hallacy, Christopher Hesse, Daniel M. Ziegler, Dario Amodei, Heewoo Jun, Jacob Jackson, Jared Kaplan, John Schulman, Mark Chen, Mor Katz, Nick Ryder, Prafulla Dhariwal, Sam McCandlish, Scott Gray, Tom B. Brown, Tom Henighan","cross_cats":["cs.CL","cs.CV"],"headline":"Autoregressive Transformers improve via power-law-plus-constant scaling laws for cross-entropy loss across image, video, text-image, and math domains, with optimal size depending on compute through nearly universal exponents.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-28T02:17:24Z","title":"Scaling Laws for Autoregressive Generative Modeling"},"references":{"count":29,"internal_anchors":5,"resolved_work":29,"sample":[{"cited_arxiv_id":"2005.14165","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Language Models are Few-Shot Learners","work_id":"214732c0-2edd-44a0-af9e-28184a2b8279","year":2020},{"cited_arxiv_id":"1904.10509","doi":"","is_internal_anchor":true,"ref_index":2,"title":"Generating Long Sequences with Sparse Transformers","work_id":"c5b81688-45ee-4a9a-b095-e6290f45cb6c","year":1904},{"cited_arxiv_id":"1707.08819","doi":"","is_internal_anchor":false,"ref_index":3,"title":"A Downsampled Variant of ImageNet as an Alternative to the CIFAR datasets","work_id":"b28da76e-30a0-4103-a956-55ca47f3e243","year":2017},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Generative pretraining from pixels","work_id":"e47419f8-22d5-401d-899d-bcc702f9885f","year":2020},{"cited_arxiv_id":"2005.00341","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Jukebox: A Generative Model for Music","work_id":"aeb5293c-aee5-4e2d-ae57-668eb930893e","year":2020}],"snapshot_sha256":"9d9f4af64c1b8a1da26c8877cbed0de0ddc8513600fc74125780a7db5ff4228d"},"source":{"id":"2010.14701","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-13T07:44:52.955019Z","id":"c673df1b-51cd-40ed-b1f0-dbcd8b110e14","model_set":{"reader":"grok-4.3"},"one_line_summary":"Autoregressive transformers follow power-law scaling laws for cross-entropy loss with nearly universal exponents relating optimal model size to compute budget across four domains.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Autoregressive Transformers improve via power-law-plus-constant scaling laws for cross-entropy loss across image, video, text-image, and math domains, with optimal size depending on compute through nearly universal exponents.","strongest_claim":"The optimal model size depends on the compute budget through a power-law, with exponents that are nearly universal across all data domains.","weakest_assumption":"That the power-law-plus-constant functional form continues to hold outside the range of model sizes and compute budgets actually measured in the experiments."}},"verdict_id":"c673df1b-51cd-40ed-b1f0-dbcd8b110e14"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e3a6b78fa7e20ef099836eac0605fcea0fe92a578e5d31a1e7031310ea844122","target":"record","created_at":"2026-07-05T01:49:38Z","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":"50db097e855a42be45810223f70bda72257bbf6cdf78c3452587fe3b8c741924","cross_cats_sorted":["cs.CL","cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-28T02:17:24Z","title_canon_sha256":"1f943d19ab5e7a38eac39acd563156ca4196d2797172c4b2ac8a298f40aa7643"},"schema_version":"1.0","source":{"id":"2010.14701","kind":"arxiv","version":2}},"canonical_sha256":"d3d59543716c7440ab3ea5804b8b309450156f477e95f20f00ade73f9430539c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d3d59543716c7440ab3ea5804b8b309450156f477e95f20f00ade73f9430539c","first_computed_at":"2026-07-05T01:49:38.992685Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T01:49:38.992685Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"yW1RFQKqKanIyapciTUO1sGh6fg76OGcnSTZqSgJ61WNkT6IdI/4w3Ja5uIQwTa7e2H06vmSz/dc9o8rM5jQDw==","signature_status":"signed_v1","signed_at":"2026-07-05T01:49:38.993111Z","signed_message":"canonical_sha256_bytes"},"source_id":"2010.14701","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e3a6b78fa7e20ef099836eac0605fcea0fe92a578e5d31a1e7031310ea844122","sha256:7d4cc30999b5f93b44a42ffbce876880697c6723d398d6b3496e478e12c6884a"],"state_sha256":"0504a2c6a4d24b1b9d09240efe4a973258548630bb48b8edc331e9a451da52b3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ixDCB535YkCdG9fziGu7vIEWJCUygCUPecsE/31KpY2jSrmAItHPjsNOXJYnW2nUj66jdcLGJEiPzbtLeBWnCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T13:37:15.647280Z","bundle_sha256":"a4d5eec4a2e0cf06df45efdb982677354281fa7c0a5de94f50caf22139b77b65"}}