{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:K2Z5L2JU7EFV6UOJTIKOWPRNSF","short_pith_number":"pith:K2Z5L2JU","canonical_record":{"source":{"id":"2305.06161","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-09T08:16:42Z","cross_cats_sorted":["cs.AI","cs.PL","cs.SE"],"title_canon_sha256":"14551b1917b51271c276b44d2aed4a3f2e7cc13a6f2f93edb6df9b5cd176c4e9","abstract_canon_sha256":"714a5370064fecaf2b21383101baf45c07f5ac114ccacec50881c99d1e37bdce"},"schema_version":"1.0"},"canonical_sha256":"56b3d5e934f90b5f51c99a14eb3e2d9170ace8e3a163a47210cf20965ed2cd9a","source":{"kind":"arxiv","id":"2305.06161","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2305.06161","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"arxiv_version","alias_value":"2305.06161v2","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.06161","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"pith_short_12","alias_value":"K2Z5L2JU7EFV","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"pith_short_16","alias_value":"K2Z5L2JU7EFV6UOJ","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"pith_short_8","alias_value":"K2Z5L2JU","created_at":"2026-07-05T07:23:26Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:K2Z5L2JU7EFV6UOJTIKOWPRNSF","target":"record","payload":{"canonical_record":{"source":{"id":"2305.06161","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-09T08:16:42Z","cross_cats_sorted":["cs.AI","cs.PL","cs.SE"],"title_canon_sha256":"14551b1917b51271c276b44d2aed4a3f2e7cc13a6f2f93edb6df9b5cd176c4e9","abstract_canon_sha256":"714a5370064fecaf2b21383101baf45c07f5ac114ccacec50881c99d1e37bdce"},"schema_version":"1.0"},"canonical_sha256":"56b3d5e934f90b5f51c99a14eb3e2d9170ace8e3a163a47210cf20965ed2cd9a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:23:26.065163Z","signature_b64":"nn29RrmKm4UESaABeW1GFLigZldy4J+SAYBGjQUSGy9BZKeyUQy1fqsZsYtlNLbCSLfaLapHJA6Df/7jCwP7Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"56b3d5e934f90b5f51c99a14eb3e2d9170ace8e3a163a47210cf20965ed2cd9a","last_reissued_at":"2026-07-05T07:23:26.063710Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:23:26.063710Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2305.06161","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":"2026-07-05T07:23:26Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"G9ZA14qrxdDaULLN/DQdvhKMKY2dkdCZpZjm2H9lF1dexuLSD8MLDO8h6TvuK0QioADW8R9lyrXT2h9I2idEBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T09:58:14.358188Z"},"content_sha256":"a4c3db26e3037dae49b7ad9d8b85ca77b2ca7e22f946b80fe00aae7bb061beb8","schema_version":"1.0","event_id":"sha256:a4c3db26e3037dae49b7ad9d8b85ca77b2ca7e22f946b80fe00aae7bb061beb8"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:K2Z5L2JU7EFV6UOJTIKOWPRNSF","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"StarCoder: may the source be with you!","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"A 15.5 billion parameter code model trained on a trillion tokens outperforms open multilingual alternatives and matches proprietary performance.","cross_cats":["cs.AI","cs.PL","cs.SE"],"primary_cat":"cs.CL","authors_text":"Alex Gu, Arjun Guha, Armel Zebaze, Benjamin Lipkin, Brendan Dolan-Gavitt, Carlos Mu\\~noz Ferrandis, Carolyn Jane Anderson, Chenghao Mou, Christopher Akiki, Claire Schlesinger, Daniel Fried, Danish Contractor, Denis Kocetkov, Dmitry Abulkhanov, Dzmitry Bahdanau, Evgenii Zheltonozhskii, Fedor Zhdanov, Hailey Schoelkopf, Harm de Vries, Jan Ebert, Jason Stillerman, Jennifer Ding, Jennifer Robinson, Jenny Chim, Jia Li, Jian Zhu, Jo\\~ao Monteiro, Joel Lamy-Poirier, Leandro Von Werra, Logesh Kumar Umapathi, Loubna Ben Allal, Manan Dey, Manuel Romero, Marc Marone, Marco Zocca, Maxim Kunakov, Mayank Mishra, Ming-Ho Yee, Mishig Davaadorj, Muhtasham Oblokulov, Nadav Timor, Nicholas Meade, Nicolas Gontier, Niklas Muennighoff, Nour Fahmy, Oleh Shliazhko, Olivier Dehaene, Paulo Villegas, Qian Liu, Raymond Li, Rudra Murthy, Sasha Luccioni, Sean Hughes, Siva Reddy, Siva Sankalp Patel, Swayam Singh, Terry Yue Zhuo, Thomas Wang, Thomas Wolf, Tony Lee, Tri Dao, Urvashi Bhattacharyya, Wenhao Yu, Yacine Jernite, Yangtian Zi, Zhihan Zhang, Zhiruo Wang","submitted_at":"2023-05-09T08:16:42Z","abstract_excerpt":"The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform th"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40% pass@1 on HumanEval, and still retains its performance on other programming languages.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The chosen benchmarks and prompting methods accurately measure real-world code generation utility across languages and tasks without hidden advantages from training data overlap or evaluation setup.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"StarCoderBase matches or beats OpenAI's code-cushman-001 on multi-language code benchmarks; the Python-fine-tuned StarCoder reaches 40% pass@1 on HumanEval while retaining other-language performance.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A 15.5 billion parameter code model trained on a trillion tokens outperforms open multilingual alternatives and matches proprietary performance.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8738eb652bbab90e10a922a7ee9f337c8b86ad581778bcf9f195f21d890fee9b"},"source":{"id":"2305.06161","kind":"arxiv","version":2},"verdict":{"id":"bae1d772-7f23-491f-b750-6d8ed95f9b8b","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T23:27:14.771914Z","strongest_claim":"StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40% pass@1 on HumanEval, and still retains its performance on other programming languages.","one_line_summary":"StarCoderBase matches or beats OpenAI's code-cushman-001 on multi-language code benchmarks; the Python-fine-tuned StarCoder reaches 40% pass@1 on HumanEval while retaining other-language performance.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The chosen benchmarks and prompting methods accurately measure real-world code generation utility across languages and tasks without hidden advantages from training data overlap or evaluation setup.","pith_extraction_headline":"A 15.5 billion parameter code model trained on a trillion tokens outperforms open multilingual alternatives and matches proprietary performance."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2305.06161/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":243,"sample":[{"doi":"","year":2021,"title":"Unified pre-training for program understanding and generation","work_id":"98d15866-1af7-4c5b-8034-96dab7ee1bf7","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Stability AI et al","work_id":"72b1345f-b7af-4714-9ae5-2e67dcfea69c","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1109/tpami.1983.4767370","year":1983,"title":"A maximum likelihood approach to continuous speech recognition","work_id":"6de51f5a-babe-4faa-a47d-e4d8bcf0097a","ref_index":7,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"ChatGPT accessible again in Italy","work_id":"3773edf0-3396-44bb-8fe7-3ae64d84c152","ref_index":9,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"A framework for the evaluation of code generation models","work_id":"a0cae6e0-1429-4088-99d6-4f3f703c80c4","ref_index":10,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":243,"snapshot_sha256":"786cdc2e35e8bd31555f16d69a70e3b406d6cd33b98e989e0f28a71b99fac870","internal_anchors":41},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"bae1d772-7f23-491f-b750-6d8ed95f9b8b"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T07:23:26Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/0XHZmjERLycyicFsw8Io4QYC6ZA3kh4WNhIHXihg+8RIsL3T7EU+rbP2/VCP7CQAyccjefgzeStamOManV5Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T09:58:14.359167Z"},"content_sha256":"d0a703329b6b79655d9d664a131f69efa0ef97eef17e266fd280b0eb5c4fa6e9","schema_version":"1.0","event_id":"sha256:d0a703329b6b79655d9d664a131f69efa0ef97eef17e266fd280b0eb5c4fa6e9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF/bundle.json","state_url":"https://pith.science/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF/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":"2026-08-05T09:58:14Z","links":{"resolver":"https://pith.science/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF","bundle":"https://pith.science/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF/bundle.json","state":"https://pith.science/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF/state.json","well_known_bundle":"https://pith.science/.well-known/pith/K2Z5L2JU7EFV6UOJTIKOWPRNSF/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:K2Z5L2JU7EFV6UOJTIKOWPRNSF","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":"714a5370064fecaf2b21383101baf45c07f5ac114ccacec50881c99d1e37bdce","cross_cats_sorted":["cs.AI","cs.PL","cs.SE"],"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-09T08:16:42Z","title_canon_sha256":"14551b1917b51271c276b44d2aed4a3f2e7cc13a6f2f93edb6df9b5cd176c4e9"},"schema_version":"1.0","source":{"id":"2305.06161","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2305.06161","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"arxiv_version","alias_value":"2305.06161v2","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.06161","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"pith_short_12","alias_value":"K2Z5L2JU7EFV","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"pith_short_16","alias_value":"K2Z5L2JU7EFV6UOJ","created_at":"2026-07-05T07:23:26Z"},{"alias_kind":"pith_short_8","alias_value":"K2Z5L2JU","created_at":"2026-07-05T07:23:26Z"}],"graph_snapshots":[{"event_id":"sha256:d0a703329b6b79655d9d664a131f69efa0ef97eef17e266fd280b0eb5c4fa6e9","target":"graph","created_at":"2026-07-05T07:23:26Z","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":"StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40% pass@1 on HumanEval, and still retains its performance on other programming languages."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The chosen benchmarks and prompting methods accurately measure real-world code generation utility across languages and tasks without hidden advantages from training data overlap or evaluation setup."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"StarCoderBase matches or beats OpenAI's code-cushman-001 on multi-language code benchmarks; the Python-fine-tuned StarCoder reaches 40% pass@1 on HumanEval while retaining other-language performance."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A 15.5 billion parameter code model trained on a trillion tokens outperforms open multilingual alternatives and matches proprietary performance."}],"snapshot_sha256":"8738eb652bbab90e10a922a7ee9f337c8b86ad581778bcf9f195f21d890fee9b"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2305.06161/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The BigCode community, an open-scientific collaboration working on the responsible development of Large Language Models for Code (Code LLMs), introduces StarCoder and StarCoderBase: 15.5B parameter models with 8K context length, infilling capabilities and fast large-batch inference enabled by multi-query attention. StarCoderBase is trained on 1 trillion tokens sourced from The Stack, a large collection of permissively licensed GitHub repositories with inspection tools and an opt-out process. We fine-tuned StarCoderBase on 35B Python tokens, resulting in the creation of StarCoder. We perform th","authors_text":"Alex Gu, Arjun Guha, Armel Zebaze, Benjamin Lipkin, Brendan Dolan-Gavitt, Carlos Mu\\~noz Ferrandis, Carolyn Jane Anderson, Chenghao Mou, Christopher Akiki, Claire Schlesinger, Daniel Fried, Danish Contractor, Denis Kocetkov, Dmitry Abulkhanov, Dzmitry Bahdanau, Evgenii Zheltonozhskii, Fedor Zhdanov, Hailey Schoelkopf, Harm de Vries, Jan Ebert, Jason Stillerman, Jennifer Ding, Jennifer Robinson, Jenny Chim, Jia Li, Jian Zhu, Jo\\~ao Monteiro, Joel Lamy-Poirier, Leandro Von Werra, Logesh Kumar Umapathi, Loubna Ben Allal, Manan Dey, Manuel Romero, Marc Marone, Marco Zocca, Maxim Kunakov, Mayank Mishra, Ming-Ho Yee, Mishig Davaadorj, Muhtasham Oblokulov, Nadav Timor, Nicholas Meade, Nicolas Gontier, Niklas Muennighoff, Nour Fahmy, Oleh Shliazhko, Olivier Dehaene, Paulo Villegas, Qian Liu, Raymond Li, Rudra Murthy, Sasha Luccioni, Sean Hughes, Siva Reddy, Siva Sankalp Patel, Swayam Singh, Terry Yue Zhuo, Thomas Wang, Thomas Wolf, Tony Lee, Tri Dao, Urvashi Bhattacharyya, Wenhao Yu, Yacine Jernite, Yangtian Zi, Zhihan Zhang, Zhiruo Wang","cross_cats":["cs.AI","cs.PL","cs.SE"],"headline":"A 15.5 billion parameter code model trained on a trillion tokens outperforms open multilingual alternatives and matches proprietary performance.","license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-09T08:16:42Z","title":"StarCoder: may the source be with you!"},"references":{"count":243,"internal_anchors":41,"resolved_work":243,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Unified pre-training for program understanding and generation","work_id":"98d15866-1af7-4c5b-8034-96dab7ee1bf7","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Stability AI et al","work_id":"72b1345f-b7af-4714-9ae5-2e67dcfea69c","year":2023},{"cited_arxiv_id":"","doi":"10.1109/tpami.1983.4767370","is_internal_anchor":false,"ref_index":7,"title":"A maximum likelihood approach to continuous speech recognition","work_id":"6de51f5a-babe-4faa-a47d-e4d8bcf0097a","year":1983},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":9,"title":"ChatGPT accessible again in Italy","work_id":"3773edf0-3396-44bb-8fe7-3ae64d84c152","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":10,"title":"A framework for the evaluation of code generation models","work_id":"a0cae6e0-1429-4088-99d6-4f3f703c80c4","year":2022}],"snapshot_sha256":"786cdc2e35e8bd31555f16d69a70e3b406d6cd33b98e989e0f28a71b99fac870"},"source":{"id":"2305.06161","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-10T23:27:14.771914Z","id":"bae1d772-7f23-491f-b750-6d8ed95f9b8b","model_set":{"reader":"grok-4.3"},"one_line_summary":"StarCoderBase matches or beats OpenAI's code-cushman-001 on multi-language code benchmarks; the Python-fine-tuned StarCoder reaches 40% pass@1 on HumanEval while retaining other-language performance.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A 15.5 billion parameter code model trained on a trillion tokens outperforms open multilingual alternatives and matches proprietary performance.","strongest_claim":"StarCoderBase outperforms every open Code LLM that supports multiple programming languages and matches or outperforms the OpenAI code-cushman-001 model. Furthermore, StarCoder outperforms every model that is fine-tuned on Python, can be prompted to achieve 40% pass@1 on HumanEval, and still retains its performance on other programming languages.","weakest_assumption":"The chosen benchmarks and prompting methods accurately measure real-world code generation utility across languages and tasks without hidden advantages from training data overlap or evaluation setup."}},"verdict_id":"bae1d772-7f23-491f-b750-6d8ed95f9b8b"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:a4c3db26e3037dae49b7ad9d8b85ca77b2ca7e22f946b80fe00aae7bb061beb8","target":"record","created_at":"2026-07-05T07:23:26Z","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":"714a5370064fecaf2b21383101baf45c07f5ac114ccacec50881c99d1e37bdce","cross_cats_sorted":["cs.AI","cs.PL","cs.SE"],"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-09T08:16:42Z","title_canon_sha256":"14551b1917b51271c276b44d2aed4a3f2e7cc13a6f2f93edb6df9b5cd176c4e9"},"schema_version":"1.0","source":{"id":"2305.06161","kind":"arxiv","version":2}},"canonical_sha256":"56b3d5e934f90b5f51c99a14eb3e2d9170ace8e3a163a47210cf20965ed2cd9a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"56b3d5e934f90b5f51c99a14eb3e2d9170ace8e3a163a47210cf20965ed2cd9a","first_computed_at":"2026-07-05T07:23:26.063710Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T07:23:26.063710Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"nn29RrmKm4UESaABeW1GFLigZldy4J+SAYBGjQUSGy9BZKeyUQy1fqsZsYtlNLbCSLfaLapHJA6Df/7jCwP7Cg==","signature_status":"signed_v1","signed_at":"2026-07-05T07:23:26.065163Z","signed_message":"canonical_sha256_bytes"},"source_id":"2305.06161","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a4c3db26e3037dae49b7ad9d8b85ca77b2ca7e22f946b80fe00aae7bb061beb8","sha256:d0a703329b6b79655d9d664a131f69efa0ef97eef17e266fd280b0eb5c4fa6e9"],"state_sha256":"a5ffe634495aa8a3c05e302422ff05463bb250f5f862e00ac3258283332e7fa6"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d90bpuSjH7BwZTKH/gBuUvQsgIDfAnHeSktVtP2L4WWaBVIMY8A+75AG1xPefkIuduGY6YxpufPmrxr9VzoDBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T09:58:14.367326Z","bundle_sha256":"0f0351e120cabdd2b68e217b6c5529cd0a75f9957c4f9f49e6c0f8cc2a2486ee"}}