{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:YO7BNP7KSPEI4VHTBGXHEIZFOV","short_pith_number":"pith:YO7BNP7K","canonical_record":{"source":{"id":"2504.05118","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-07T14:21:11Z","cross_cats_sorted":[],"title_canon_sha256":"921ebab5f42a1de740eb2f28f92f75e2ece38fbd3f5bdc26faf62024af2faaa4","abstract_canon_sha256":"909372adb6fa23641d5a19532e8c0623e35edb8a37a9d86605950495e7ec8907"},"schema_version":"1.0"},"canonical_sha256":"c3be16bfea93c88e54f309ae722325756e7f60b3303555645831315ff13d53be","source":{"kind":"arxiv","id":"2504.05118","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.05118","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"arxiv_version","alias_value":"2504.05118v3","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.05118","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"pith_short_12","alias_value":"YO7BNP7KSPEI","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"pith_short_16","alias_value":"YO7BNP7KSPEI4VHT","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"pith_short_8","alias_value":"YO7BNP7K","created_at":"2026-07-05T10:47:38Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:YO7BNP7KSPEI4VHTBGXHEIZFOV","target":"record","payload":{"canonical_record":{"source":{"id":"2504.05118","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-07T14:21:11Z","cross_cats_sorted":[],"title_canon_sha256":"921ebab5f42a1de740eb2f28f92f75e2ece38fbd3f5bdc26faf62024af2faaa4","abstract_canon_sha256":"909372adb6fa23641d5a19532e8c0623e35edb8a37a9d86605950495e7ec8907"},"schema_version":"1.0"},"canonical_sha256":"c3be16bfea93c88e54f309ae722325756e7f60b3303555645831315ff13d53be","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:47:38.878636Z","signature_b64":"kDoYsEuFpE37bPfeE+BO8VPcWJLM8jyTTWXUgHBnT6R+L1Za7beTYyJQic6XzC6Vkpyf6ebdGs8soxZt1ttSAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c3be16bfea93c88e54f309ae722325756e7f60b3303555645831315ff13d53be","last_reissued_at":"2026-07-05T10:47:38.878134Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:47:38.878134Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.05118","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":"2026-07-05T10:47:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"2VIlEdhRZZObog51a4v/PdtCAYQl2dnz5/AiayDdvljH5d1raZO8pU1X2UismktD+3LQFthY0l2tTSsMDVkiCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T05:31:12.590181Z"},"content_sha256":"b56ba57345cfb0cffe01227f4ab04833ac5d87e92957e8c54bd9bcda1d7126c2","schema_version":"1.0","event_id":"sha256:b56ba57345cfb0cffe01227f4ab04833ac5d87e92957e8c54bd9bcda1d7126c2"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:YO7BNP7KSPEI4VHTBGXHEIZFOV","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Tasks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"VAPO reaches 60.4 on AIME 2024 by fixing value bias, variable lengths, and sparse rewards in RL for reasoning.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Bole Ma, Chengyi Wang, Chi Zhang, Gaohong Liu, Haibin Lin, Hang Zhu, Jiaze Chen, Juncai Liu, Lingjun Liu, Lin Yan, Mingxuan Wang, Mofan Zhang, Qiying Yu, Ruofei Zhu, Ru Zhang, Tiantian Fan, Wang Zhang, Wenyuan Xu, Xiangpeng Wei, Xiangyu Yu, Xiaochen Zuo, Xin Liu, Yonghui Wu, Yufeng Yuan, Yu Yue, Zhengyin Du, Zhiqi Lin","submitted_at":"2025-04-07T14:21:11Z","abstract_excerpt":"We present VAPO, Value-based Augmented Proximal Policy Optimization framework for reasoning models., a novel framework tailored for reasoning models within the value-based paradigm. Benchmarked the AIME 2024 dataset, VAPO, built on the Qwen 32B pre-trained model, attains a state-of-the-art score of $\\mathbf{60.4}$. In direct comparison under identical experimental settings, VAPO outperforms the previously reported results of DeepSeek-R1-Zero-Qwen-32B and DAPO by more than 10 points. The training process of VAPO stands out for its stability and efficiency. It reaches state-of-the-art performanc"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Benchmarked the AIME 2024 dataset, VAPO, built on the Qwen 32B pre-trained model, attains a state-of-the-art score of 60.4. In direct comparison under identical experimental settings, VAPO outperforms the previously reported results of DeepSeek-R1-Zero-Qwen-32B and DAPO by more than 10 points. The training process of VAPO stands out for its stability and efficiency. It reaches state-of-the-art performance within a mere 5,000 steps. Moreover, across multiple independent runs, no training crashes occur.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the performance improvements and stability are due to the specific VAPO design choices rather than unreported differences in training data, hyperparameters, model initialization, or evaluation protocols, as the abstract provides no details on these controls.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"VAPO achieves 60.4 on AIME 2024 with Qwen 32B, outperforming prior methods by over 10 points through targeted fixes for value bias, sequence length variation, and sparse rewards.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"VAPO reaches 60.4 on AIME 2024 by fixing value bias, variable lengths, and sparse rewards in RL for reasoning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"bf6a613e6540a8ede942b81eb6e4290de04a0c882544d9a6f57837179e4c1723"},"source":{"id":"2504.05118","kind":"arxiv","version":3},"verdict":{"id":"81318f0a-79df-49f8-a70f-5a0c79ad6d49","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-13T09:30:15.431236Z","strongest_claim":"Benchmarked the AIME 2024 dataset, VAPO, built on the Qwen 32B pre-trained model, attains a state-of-the-art score of 60.4. In direct comparison under identical experimental settings, VAPO outperforms the previously reported results of DeepSeek-R1-Zero-Qwen-32B and DAPO by more than 10 points. The training process of VAPO stands out for its stability and efficiency. It reaches state-of-the-art performance within a mere 5,000 steps. Moreover, across multiple independent runs, no training crashes occur.","one_line_summary":"VAPO achieves 60.4 on AIME 2024 with Qwen 32B, outperforming prior methods by over 10 points through targeted fixes for value bias, sequence length variation, and sparse rewards.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the performance improvements and stability are due to the specific VAPO design choices rather than unreported differences in training data, hyperparameters, model initialization, or evaluation protocols, as the abstract provides no details on these controls.","pith_extraction_headline":"VAPO reaches 60.4 on AIME 2024 by fixing value bias, variable lengths, and sparse rewards in RL for reasoning."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2504.05118/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":30,"sample":[{"doi":"","year":2024,"title":"Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs","work_id":"7bb8f9ec-1241-4472-a4fa-c636c6d79892","ref_index":1,"cited_arxiv_id":"2402.14740","is_internal_anchor":true},{"doi":"","year":2024,"title":"Claude 3.5 sonnet","work_id":"e2a59025-aef0-493f-83c4-f1608cf2b5f0","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1901,"title":"Language models are few-shot learners","work_id":"1109e15b-0e77-4d1f-9248-ed7317a8400c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Palm: Scaling language modeling with pathways","work_id":"815cff7d-cce6-4360-ba9e-2c1f9f716e29","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Gemini 2.0 flash thinking","work_id":"80f81270-beb6-4b89-835f-428d42b2ca53","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":30,"snapshot_sha256":"1bbffa45251d5e0dde99678451c137063516b8894ccf2c648fff56ac3cf8b864","internal_anchors":13},"formal_canon":{"evidence_count":2,"snapshot_sha256":"7d9cbdea1c825611d475da16a1fff94d645b0b5ea09dbf5ecb845bb2ddfc8c68"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"81318f0a-79df-49f8-a70f-5a0c79ad6d49"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:47:38Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"nw+46Ip5yNplNE/duaU+wGJAYPc9e/fUemq6Mxv7WoTuAXawUFy17CKfNRFkpCjJKJTU4qxoHVteU4XVNEShBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-11T05:31:12.591051Z"},"content_sha256":"3d19632d6cf63b52ded2e693cae3f74a3168d7740ad7453e357c585513e1d9d8","schema_version":"1.0","event_id":"sha256:3d19632d6cf63b52ded2e693cae3f74a3168d7740ad7453e357c585513e1d9d8"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV/bundle.json","state_url":"https://pith.science/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV/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":"2026-08-11T05:31:12Z","links":{"resolver":"https://pith.science/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV","bundle":"https://pith.science/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV/bundle.json","state":"https://pith.science/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV/state.json","well_known_bundle":"https://pith.science/.well-known/pith/YO7BNP7KSPEI4VHTBGXHEIZFOV/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:YO7BNP7KSPEI4VHTBGXHEIZFOV","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":"909372adb6fa23641d5a19532e8c0623e35edb8a37a9d86605950495e7ec8907","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-07T14:21:11Z","title_canon_sha256":"921ebab5f42a1de740eb2f28f92f75e2ece38fbd3f5bdc26faf62024af2faaa4"},"schema_version":"1.0","source":{"id":"2504.05118","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.05118","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"arxiv_version","alias_value":"2504.05118v3","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.05118","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"pith_short_12","alias_value":"YO7BNP7KSPEI","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"pith_short_16","alias_value":"YO7BNP7KSPEI4VHT","created_at":"2026-07-05T10:47:38Z"},{"alias_kind":"pith_short_8","alias_value":"YO7BNP7K","created_at":"2026-07-05T10:47:38Z"}],"graph_snapshots":[{"event_id":"sha256:3d19632d6cf63b52ded2e693cae3f74a3168d7740ad7453e357c585513e1d9d8","target":"graph","created_at":"2026-07-05T10:47:38Z","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":"Benchmarked the AIME 2024 dataset, VAPO, built on the Qwen 32B pre-trained model, attains a state-of-the-art score of 60.4. In direct comparison under identical experimental settings, VAPO outperforms the previously reported results of DeepSeek-R1-Zero-Qwen-32B and DAPO by more than 10 points. The training process of VAPO stands out for its stability and efficiency. It reaches state-of-the-art performance within a mere 5,000 steps. Moreover, across multiple independent runs, no training crashes occur."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the performance improvements and stability are due to the specific VAPO design choices rather than unreported differences in training data, hyperparameters, model initialization, or evaluation protocols, as the abstract provides no details on these controls."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"VAPO achieves 60.4 on AIME 2024 with Qwen 32B, outperforming prior methods by over 10 points through targeted fixes for value bias, sequence length variation, and sparse rewards."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"VAPO reaches 60.4 on AIME 2024 by fixing value bias, variable lengths, and sparse rewards in RL for reasoning."}],"snapshot_sha256":"bf6a613e6540a8ede942b81eb6e4290de04a0c882544d9a6f57837179e4c1723"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"7d9cbdea1c825611d475da16a1fff94d645b0b5ea09dbf5ecb845bb2ddfc8c68"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2504.05118/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We present VAPO, Value-based Augmented Proximal Policy Optimization framework for reasoning models., a novel framework tailored for reasoning models within the value-based paradigm. Benchmarked the AIME 2024 dataset, VAPO, built on the Qwen 32B pre-trained model, attains a state-of-the-art score of $\\mathbf{60.4}$. In direct comparison under identical experimental settings, VAPO outperforms the previously reported results of DeepSeek-R1-Zero-Qwen-32B and DAPO by more than 10 points. The training process of VAPO stands out for its stability and efficiency. It reaches state-of-the-art performanc","authors_text":"Bole Ma, Chengyi Wang, Chi Zhang, Gaohong Liu, Haibin Lin, Hang Zhu, Jiaze Chen, Juncai Liu, Lingjun Liu, Lin Yan, Mingxuan Wang, Mofan Zhang, Qiying Yu, Ruofei Zhu, Ru Zhang, Tiantian Fan, Wang Zhang, Wenyuan Xu, Xiangpeng Wei, Xiangyu Yu, Xiaochen Zuo, Xin Liu, Yonghui Wu, Yufeng Yuan, Yu Yue, Zhengyin Du, Zhiqi Lin","cross_cats":[],"headline":"VAPO reaches 60.4 on AIME 2024 by fixing value bias, variable lengths, and sparse rewards in RL for reasoning.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-07T14:21:11Z","title":"VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Tasks"},"references":{"count":30,"internal_anchors":13,"resolved_work":30,"sample":[{"cited_arxiv_id":"2402.14740","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMs","work_id":"7bb8f9ec-1241-4472-a4fa-c636c6d79892","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Claude 3.5 sonnet","work_id":"e2a59025-aef0-493f-83c4-f1608cf2b5f0","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Language models are few-shot learners","work_id":"1109e15b-0e77-4d1f-9248-ed7317a8400c","year":1901},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Palm: Scaling language modeling with pathways","work_id":"815cff7d-cce6-4360-ba9e-2c1f9f716e29","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Gemini 2.0 flash thinking","work_id":"80f81270-beb6-4b89-835f-428d42b2ca53","year":2024}],"snapshot_sha256":"1bbffa45251d5e0dde99678451c137063516b8894ccf2c648fff56ac3cf8b864"},"source":{"id":"2504.05118","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-13T09:30:15.431236Z","id":"81318f0a-79df-49f8-a70f-5a0c79ad6d49","model_set":{"reader":"grok-4.3"},"one_line_summary":"VAPO achieves 60.4 on AIME 2024 with Qwen 32B, outperforming prior methods by over 10 points through targeted fixes for value bias, sequence length variation, and sparse rewards.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"VAPO reaches 60.4 on AIME 2024 by fixing value bias, variable lengths, and sparse rewards in RL for reasoning.","strongest_claim":"Benchmarked the AIME 2024 dataset, VAPO, built on the Qwen 32B pre-trained model, attains a state-of-the-art score of 60.4. In direct comparison under identical experimental settings, VAPO outperforms the previously reported results of DeepSeek-R1-Zero-Qwen-32B and DAPO by more than 10 points. The training process of VAPO stands out for its stability and efficiency. It reaches state-of-the-art performance within a mere 5,000 steps. Moreover, across multiple independent runs, no training crashes occur.","weakest_assumption":"That the performance improvements and stability are due to the specific VAPO design choices rather than unreported differences in training data, hyperparameters, model initialization, or evaluation protocols, as the abstract provides no details on these controls."}},"verdict_id":"81318f0a-79df-49f8-a70f-5a0c79ad6d49"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:b56ba57345cfb0cffe01227f4ab04833ac5d87e92957e8c54bd9bcda1d7126c2","target":"record","created_at":"2026-07-05T10:47:38Z","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":"909372adb6fa23641d5a19532e8c0623e35edb8a37a9d86605950495e7ec8907","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2025-04-07T14:21:11Z","title_canon_sha256":"921ebab5f42a1de740eb2f28f92f75e2ece38fbd3f5bdc26faf62024af2faaa4"},"schema_version":"1.0","source":{"id":"2504.05118","kind":"arxiv","version":3}},"canonical_sha256":"c3be16bfea93c88e54f309ae722325756e7f60b3303555645831315ff13d53be","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c3be16bfea93c88e54f309ae722325756e7f60b3303555645831315ff13d53be","first_computed_at":"2026-07-05T10:47:38.878134Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:47:38.878134Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"kDoYsEuFpE37bPfeE+BO8VPcWJLM8jyTTWXUgHBnT6R+L1Za7beTYyJQic6XzC6Vkpyf6ebdGs8soxZt1ttSAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:47:38.878636Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.05118","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b56ba57345cfb0cffe01227f4ab04833ac5d87e92957e8c54bd9bcda1d7126c2","sha256:3d19632d6cf63b52ded2e693cae3f74a3168d7740ad7453e357c585513e1d9d8"],"state_sha256":"3a7ecaf5a37d2ef0f02f51b17a6e5b44142348429a11d5ca142bcffa59d8918a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LzH9z4AMfCxCZdjUMUsK63pP+f+6Vj+n3xOWrKk8IeXF4heigR7OlZ+2u3EFvy5BsqTYSSLHIIPc3ZFVq13fDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-11T05:31:12.596952Z","bundle_sha256":"618c77a60d8997c5dd63471be3e6ca66a53cabf103ac83aacdbdbe3c2509ba0c"}}