{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:BFFGISG2S7M2YH25DA67B6UIJM","short_pith_number":"pith:BFFGISG2","canonical_record":{"source":{"id":"2402.17152","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-27T02:37:37Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"80b689fef0442192436239259187c0285189c408eb8b65ccb0e18827fc933a69","abstract_canon_sha256":"7307ede0c99a0d3921300dae34e93bd3ac74e28fce8ac10f41bc1d0b98ff986e"},"schema_version":"1.0"},"canonical_sha256":"094a6448da97d9ac1f5d183df0fa884b2ca812508c8387cc476032d782f552d9","source":{"kind":"arxiv","id":"2402.17152","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.17152","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"arxiv_version","alias_value":"2402.17152v3","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.17152","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"pith_short_12","alias_value":"BFFGISG2S7M2","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"pith_short_16","alias_value":"BFFGISG2S7M2YH25","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"pith_short_8","alias_value":"BFFGISG2","created_at":"2026-07-05T08:15:45Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:BFFGISG2S7M2YH25DA67B6UIJM","target":"record","payload":{"canonical_record":{"source":{"id":"2402.17152","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-27T02:37:37Z","cross_cats_sorted":["cs.IR"],"title_canon_sha256":"80b689fef0442192436239259187c0285189c408eb8b65ccb0e18827fc933a69","abstract_canon_sha256":"7307ede0c99a0d3921300dae34e93bd3ac74e28fce8ac10f41bc1d0b98ff986e"},"schema_version":"1.0"},"canonical_sha256":"094a6448da97d9ac1f5d183df0fa884b2ca812508c8387cc476032d782f552d9","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:15:45.769053Z","signature_b64":"YtSh7LSddlYAQkvpPYAQYrKhJs10/JkZUZRdt8ciBzxXyTqPa/BJ60/Q4tvX7SNrJdDOkkuwF0fVg8w6C8QHBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"094a6448da97d9ac1f5d183df0fa884b2ca812508c8387cc476032d782f552d9","last_reissued_at":"2026-07-05T08:15:45.768556Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:15:45.768556Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2402.17152","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":"2026-07-05T08:15:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"m+TehCwgkMiDO0uXyMdvwgjool7/VM/PnvZ0aQcoFIK+peW7qEHFZzgogjlj+cugCT5YXlblWT6lBY3IXrwqAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-01T04:50:15.065584Z"},"content_sha256":"ed75f238680a0389024ca87d2478bf610a88a2e8fd4b0a58fbdd5e2797480ab0","schema_version":"1.0","event_id":"sha256:ed75f238680a0389024ca87d2478bf610a88a2e8fd4b0a58fbdd5e2797480ab0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:BFFGISG2S7M2YH25DA67B6UIJM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"Reformulating recommendations as generative sequential transduction with the HSTU architecture lets models scale to 1.5 trillion parameters following power-law improvements in compute.","cross_cats":["cs.IR"],"primary_cat":"cs.LG","authors_text":"Fangda Gu, Jiaqi Zhai, Leon Gao, Lucy Liao, Michael He, Rui Li, Xing Liu, Xuan Cao, Yinghai Lu, Yueming Wang, Yu Shi, Zhaojie Gong","submitted_at":"2024-02-27T02:37:37Z","abstract_excerpt":"Large-scale recommendation systems are characterized by their reliance on high cardinality, heterogeneous features and the need to handle tens of billions of user actions on a daily basis. Despite being trained on huge volume of data with thousands of features, most Deep Learning Recommendation Models (DLRMs) in industry fail to scale with compute.\n  Inspired by success achieved by Transformers in language and vision domains, we revisit fundamental design choices in recommendation systems. We reformulate recommendation problems as sequential transduction tasks within a generative modeling fram"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"HSTU-based Generative Recommenders, with 1.5 trillion parameters, improve metrics in online A/B tests by 12.4% and have been deployed on multiple surfaces of a large internet platform with billions of users. More importantly, the model quality of Generative Recommenders empirically scales as a power-law of training compute across three orders of magnitude, up to GPT-3/LLaMa-2 scale.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the reformulation of recommendation as generative sequential transduction with the HSTU architecture captures the core dynamics of non-stationary, high-cardinality user behavior without introducing artifacts that would not appear in standard DLRM baselines under identical training regimes.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"HSTU-based generative recommenders with 1.5 trillion parameters scale as a power law with compute up to GPT-3 scale, outperform baselines by up to 65.8% NDCG, run 5-15x faster than FlashAttention2 on long sequences, and improve online A/B metrics by 12.4%.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Reformulating recommendations as generative sequential transduction with the HSTU architecture lets models scale to 1.5 trillion parameters following power-law improvements in compute.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"a57728029839fe3f192213876a9046a6dff749d304b406ad1e677cb56d21f3cb"},"source":{"id":"2402.17152","kind":"arxiv","version":3},"verdict":{"id":"e9b6e133-88db-4e90-a830-dccf807fa479","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T19:34:34.307296Z","strongest_claim":"HSTU-based Generative Recommenders, with 1.5 trillion parameters, improve metrics in online A/B tests by 12.4% and have been deployed on multiple surfaces of a large internet platform with billions of users. More importantly, the model quality of Generative Recommenders empirically scales as a power-law of training compute across three orders of magnitude, up to GPT-3/LLaMa-2 scale.","one_line_summary":"HSTU-based generative recommenders with 1.5 trillion parameters scale as a power law with compute up to GPT-3 scale, outperform baselines by up to 65.8% NDCG, run 5-15x faster than FlashAttention2 on long sequences, and improve online A/B metrics by 12.4%.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the reformulation of recommendation as generative sequential transduction with the HSTU architecture captures the core dynamics of non-stationary, high-cardinality user behavior without introducing artifacts that would not appear in standard DLRM baselines under identical training regimes.","pith_extraction_headline":"Reformulating recommendations as generative sequential transduction with the HSTU architecture lets models scale to 1.5 trillion parameters following power-law improvements in compute."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2402.17152/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":143,"sample":[{"doi":"","year":null,"title":"and Ermon, Stefano and Rudra, Atri and R","work_id":"26b2e9b3-3881-4efb-8d42-91e7b2fa79ce","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning , author=. 2023 , eprint=","work_id":"bb11d244-c97e-4e9d-9c10-6b62be133029","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"Efficiently Scaling Transformer Inference , author=. 2022 , eprint=","work_id":"f0a5e38f-f8df-4f9a-8e2c-d0dd46df96dd","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"COLD: Towards the Next Generation of Pre-Ranking System , author=. 2020 , eprint=","work_id":"1c64fe64-4f87-43b3-b06b-b02ebcb38427","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Asymmetric LSH (ALSH) for Sublinear Time Maximum Inner Product Search (MIPS) , volume =","work_id":"219a9575-77ea-4df7-9c01-58d277ce6cef","ref_index":6,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":143,"snapshot_sha256":"b00d451726d35168b1674538c776566eed625036f348aa0aa233a8c1e6dac347","internal_anchors":5},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f84dfea2274138c74250c90e9adc7726c02843db864b79747440e5d04e3fe7c1"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"e9b6e133-88db-4e90-a830-dccf807fa479"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:15:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"MnRXkEsRQn/GWXqf6+cWM/Ijco9CIjdl0nARdgB5Jp2dq8cv4L5mvqVIKDNkPtV0MJs8FATAQ1UbKNK5CCz9AQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-01T04:50:15.066585Z"},"content_sha256":"0e464037900cba43a76b0afe383cf65a36069aa30ec4c4b0de9d6c937cbc4de8","schema_version":"1.0","event_id":"sha256:0e464037900cba43a76b0afe383cf65a36069aa30ec4c4b0de9d6c937cbc4de8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/BFFGISG2S7M2YH25DA67B6UIJM/bundle.json","state_url":"https://pith.science/pith/BFFGISG2S7M2YH25DA67B6UIJM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/BFFGISG2S7M2YH25DA67B6UIJM/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":"2026-08-01T04:50:15Z","links":{"resolver":"https://pith.science/pith/BFFGISG2S7M2YH25DA67B6UIJM","bundle":"https://pith.science/pith/BFFGISG2S7M2YH25DA67B6UIJM/bundle.json","state":"https://pith.science/pith/BFFGISG2S7M2YH25DA67B6UIJM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/BFFGISG2S7M2YH25DA67B6UIJM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:BFFGISG2S7M2YH25DA67B6UIJM","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":"7307ede0c99a0d3921300dae34e93bd3ac74e28fce8ac10f41bc1d0b98ff986e","cross_cats_sorted":["cs.IR"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-27T02:37:37Z","title_canon_sha256":"80b689fef0442192436239259187c0285189c408eb8b65ccb0e18827fc933a69"},"schema_version":"1.0","source":{"id":"2402.17152","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2402.17152","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"arxiv_version","alias_value":"2402.17152v3","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.17152","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"pith_short_12","alias_value":"BFFGISG2S7M2","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"pith_short_16","alias_value":"BFFGISG2S7M2YH25","created_at":"2026-07-05T08:15:45Z"},{"alias_kind":"pith_short_8","alias_value":"BFFGISG2","created_at":"2026-07-05T08:15:45Z"}],"graph_snapshots":[{"event_id":"sha256:0e464037900cba43a76b0afe383cf65a36069aa30ec4c4b0de9d6c937cbc4de8","target":"graph","created_at":"2026-07-05T08:15:45Z","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":"HSTU-based Generative Recommenders, with 1.5 trillion parameters, improve metrics in online A/B tests by 12.4% and have been deployed on multiple surfaces of a large internet platform with billions of users. More importantly, the model quality of Generative Recommenders empirically scales as a power-law of training compute across three orders of magnitude, up to GPT-3/LLaMa-2 scale."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the reformulation of recommendation as generative sequential transduction with the HSTU architecture captures the core dynamics of non-stationary, high-cardinality user behavior without introducing artifacts that would not appear in standard DLRM baselines under identical training regimes."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"HSTU-based generative recommenders with 1.5 trillion parameters scale as a power law with compute up to GPT-3 scale, outperform baselines by up to 65.8% NDCG, run 5-15x faster than FlashAttention2 on long sequences, and improve online A/B metrics by 12.4%."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Reformulating recommendations as generative sequential transduction with the HSTU architecture lets models scale to 1.5 trillion parameters following power-law improvements in compute."}],"snapshot_sha256":"a57728029839fe3f192213876a9046a6dff749d304b406ad1e677cb56d21f3cb"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f84dfea2274138c74250c90e9adc7726c02843db864b79747440e5d04e3fe7c1"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2402.17152/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large-scale recommendation systems are characterized by their reliance on high cardinality, heterogeneous features and the need to handle tens of billions of user actions on a daily basis. Despite being trained on huge volume of data with thousands of features, most Deep Learning Recommendation Models (DLRMs) in industry fail to scale with compute.\n  Inspired by success achieved by Transformers in language and vision domains, we revisit fundamental design choices in recommendation systems. We reformulate recommendation problems as sequential transduction tasks within a generative modeling fram","authors_text":"Fangda Gu, Jiaqi Zhai, Leon Gao, Lucy Liao, Michael He, Rui Li, Xing Liu, Xuan Cao, Yinghai Lu, Yueming Wang, Yu Shi, Zhaojie Gong","cross_cats":["cs.IR"],"headline":"Reformulating recommendations as generative sequential transduction with the HSTU architecture lets models scale to 1.5 trillion parameters following power-law improvements in compute.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-27T02:37:37Z","title":"Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations"},"references":{"count":143,"internal_anchors":5,"resolved_work":143,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"and Ermon, Stefano and Rudra, Atri and R","work_id":"26b2e9b3-3881-4efb-8d42-91e7b2fa79ce","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning , author=. 2023 , eprint=","work_id":"bb11d244-c97e-4e9d-9c10-6b62be133029","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Efficiently Scaling Transformer Inference , author=. 2022 , eprint=","work_id":"f0a5e38f-f8df-4f9a-8e2c-d0dd46df96dd","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"COLD: Towards the Next Generation of Pre-Ranking System , author=. 2020 , eprint=","work_id":"1c64fe64-4f87-43b3-b06b-b02ebcb38427","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":6,"title":"Asymmetric LSH (ALSH) for Sublinear Time Maximum Inner Product Search (MIPS) , volume =","work_id":"219a9575-77ea-4df7-9c01-58d277ce6cef","year":null}],"snapshot_sha256":"b00d451726d35168b1674538c776566eed625036f348aa0aa233a8c1e6dac347"},"source":{"id":"2402.17152","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-13T19:34:34.307296Z","id":"e9b6e133-88db-4e90-a830-dccf807fa479","model_set":{"reader":"grok-4.3"},"one_line_summary":"HSTU-based generative recommenders with 1.5 trillion parameters scale as a power law with compute up to GPT-3 scale, outperform baselines by up to 65.8% NDCG, run 5-15x faster than FlashAttention2 on long sequences, and improve online A/B metrics by 12.4%.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Reformulating recommendations as generative sequential transduction with the HSTU architecture lets models scale to 1.5 trillion parameters following power-law improvements in compute.","strongest_claim":"HSTU-based Generative Recommenders, with 1.5 trillion parameters, improve metrics in online A/B tests by 12.4% and have been deployed on multiple surfaces of a large internet platform with billions of users. More importantly, the model quality of Generative Recommenders empirically scales as a power-law of training compute across three orders of magnitude, up to GPT-3/LLaMa-2 scale.","weakest_assumption":"That the reformulation of recommendation as generative sequential transduction with the HSTU architecture captures the core dynamics of non-stationary, high-cardinality user behavior without introducing artifacts that would not appear in standard DLRM baselines under identical training regimes."}},"verdict_id":"e9b6e133-88db-4e90-a830-dccf807fa479"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ed75f238680a0389024ca87d2478bf610a88a2e8fd4b0a58fbdd5e2797480ab0","target":"record","created_at":"2026-07-05T08:15:45Z","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":"7307ede0c99a0d3921300dae34e93bd3ac74e28fce8ac10f41bc1d0b98ff986e","cross_cats_sorted":["cs.IR"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-27T02:37:37Z","title_canon_sha256":"80b689fef0442192436239259187c0285189c408eb8b65ccb0e18827fc933a69"},"schema_version":"1.0","source":{"id":"2402.17152","kind":"arxiv","version":3}},"canonical_sha256":"094a6448da97d9ac1f5d183df0fa884b2ca812508c8387cc476032d782f552d9","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"094a6448da97d9ac1f5d183df0fa884b2ca812508c8387cc476032d782f552d9","first_computed_at":"2026-07-05T08:15:45.768556Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:15:45.768556Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"YtSh7LSddlYAQkvpPYAQYrKhJs10/JkZUZRdt8ciBzxXyTqPa/BJ60/Q4tvX7SNrJdDOkkuwF0fVg8w6C8QHBg==","signature_status":"signed_v1","signed_at":"2026-07-05T08:15:45.769053Z","signed_message":"canonical_sha256_bytes"},"source_id":"2402.17152","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ed75f238680a0389024ca87d2478bf610a88a2e8fd4b0a58fbdd5e2797480ab0","sha256:0e464037900cba43a76b0afe383cf65a36069aa30ec4c4b0de9d6c937cbc4de8"],"state_sha256":"699a0f93d7bf0d74ac215d951b9e4210842c75f760944dd97713ebb6157ad45b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7c2HvjgoqmpY1VKezQehr1yaI/7jcQquhu3bAeaHue6meGU5a4FnOlR1pPOA9RIU4eUWUrZhu4gptUy/YqnVAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-01T04:50:15.072259Z","bundle_sha256":"eed7766c57f50ffaecf4f5c38157dbf1a1f6123f3aabec6ba6b0516ae6182723"}}