{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:IST3XHVNJSADRSBL5L6JRTVLNT","short_pith_number":"pith:IST3XHVN","canonical_record":{"source":{"id":"2304.03277","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-04-06T17:58:09Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2da0c9d0dd1bb0ca8ff88374f47294a26d976e55871dbc5d18aa2d9f2cae193e","abstract_canon_sha256":"a5fa7e08c14cfe870725a2b0ab026cc82993bc6daed0af6062dcbe35be363a91"},"schema_version":"1.0"},"canonical_sha256":"44a7bb9ead4c8038c82beafc98ceab6ce87f1de79c8e0b9e50615d9cc1fee7af","source":{"kind":"arxiv","id":"2304.03277","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2304.03277","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"arxiv_version","alias_value":"2304.03277v1","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.03277","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"pith_short_12","alias_value":"IST3XHVNJSAD","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"IST3XHVNJSADRSBL","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"IST3XHVN","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:IST3XHVNJSADRSBL5L6JRTVLNT","target":"record","payload":{"canonical_record":{"source":{"id":"2304.03277","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-04-06T17:58:09Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2da0c9d0dd1bb0ca8ff88374f47294a26d976e55871dbc5d18aa2d9f2cae193e","abstract_canon_sha256":"a5fa7e08c14cfe870725a2b0ab026cc82993bc6daed0af6062dcbe35be363a91"},"schema_version":"1.0"},"canonical_sha256":"44a7bb9ead4c8038c82beafc98ceab6ce87f1de79c8e0b9e50615d9cc1fee7af","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:22.360701Z","signature_b64":"S+UED9yYs0+zmAMcnpwQloKfVlHgac9qhW074tDP0h0a3aGJj/IFy9dMI9WPbE3j1ApZ4w4ww/V7Q2v3N0DzAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"44a7bb9ead4c8038c82beafc98ceab6ce87f1de79c8e0b9e50615d9cc1fee7af","last_reissued_at":"2026-05-17T23:39:22.359932Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:22.359932Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2304.03277","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":"2026-05-17T23:39:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3UcOFEXgsx4x3nYVKqvkH/XNnThB1AjPXoHVHZiobTBzzjHVNIowqwfX8AUh9pepVtHOL3PsAcvevE8Q+6EfAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-04T19:34:47.700003Z"},"content_sha256":"a32cc948bf1a828856f5d88ce7603244c9939bc0ed57a9823689f514a558a216","schema_version":"1.0","event_id":"sha256:a32cc948bf1a828856f5d88ce7603244c9939bc0ed57a9823689f514a558a216"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:IST3XHVNJSADRSBL5L6JRTVLNT","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Instruction Tuning with GPT-4","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"GPT-4 generated instruction data enables LLaMA models to reach higher zero-shot performance on new tasks than earlier synthetic datasets.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Baolin Peng, Chunyuan Li, Jianfeng Gao, Michel Galley, Pengcheng He","submitted_at":"2023-04-06T17:58:09Z","abstract_excerpt":"Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by pr"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That performance gains on the chosen zero-shot tasks reflect genuine improvement in instruction following rather than artifacts of GPT-4's data generation process or evaluation choices.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"GPT-4-generated instruction data produces superior zero-shot performance in finetuned LLaMA models versus prior state-of-the-art data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"GPT-4 generated instruction data enables LLaMA models to reach higher zero-shot performance on new tasks than earlier synthetic datasets.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3e5864af4ec74bde1b4fb3b6168781c91bd2d762b48e55b99d23d6306483114d"},"source":{"id":"2304.03277","kind":"arxiv","version":1},"verdict":{"id":"7170c55c-42cb-42e8-b30e-0c41f8928e3d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T16:57:59.957977Z","strongest_claim":"Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models.","one_line_summary":"GPT-4-generated instruction data produces superior zero-shot performance in finetuned LLaMA models versus prior state-of-the-art data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That performance gains on the chosen zero-shot tasks reflect genuine improvement in instruction following rather than artifacts of GPT-4's data generation process or evaluation choices.","pith_extraction_headline":"GPT-4 generated instruction data enables LLaMA models to reach higher zero-shot performance on new tasks than earlier synthetic datasets."},"references":{"count":18,"sample":[{"doi":"","year":null,"title":"A General Language Assistant as a Laboratory for Alignment","work_id":"a43f9ea0-01be-47d5-b8ee-a1a9f73381c5","ref_index":1,"cited_arxiv_id":"2112.00861","is_internal_anchor":true},{"doi":"10.5281/zenodo.7733589","year":null,"title":"Yejin Bang, Samuel Cahyawijaya, Nayeon Lee, Wenliang Dai, Dan Su, Bryan Wilie, Holy Lovenia, Ziwei Ji, Tiezheng Yu, Willy Chung, et al","work_id":"6f218053-cca5-4fde-92aa-730589931f0c","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Constitutional AI: Harmlessness from AI Feedback","work_id":"faaaa4e0-2676-4fac-a0b4-99aef10d2095","ref_index":3,"cited_arxiv_id":"2212.08073","is_internal_anchor":true},{"doi":"10.5281/zenodo.5297715","year":1901,"title":"GPT-Neo: Large Scale Autoregressive Language Modeling with Mesh-Tensorflow , March 2021","work_id":"6c7f8a44-6f52-448c-b819-5ba82a7bbc59","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Scaling Instruction-Finetuned Language Models","work_id":"8405abb1-7558-4fdf-af24-f4c52fa77a06","ref_index":5,"cited_arxiv_id":"2210.11416","is_internal_anchor":true}],"resolved_work":18,"snapshot_sha256":"086a76bdf9860cd1a85670949cbde7831e84f7f71ab1d678f23db21772a9c2ed","internal_anchors":11},"formal_canon":{"evidence_count":1,"snapshot_sha256":"c26f16a8dcc7d1eb015e86884555556a48859d117881c5f57bf53f828be0fea0"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"7170c55c-42cb-42e8-b30e-0c41f8928e3d"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:22Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Kgs7l6Eej69YPP9JzGpwVtZphSjsScfjEbZaZq3n+O8Dl5PLSPQ7M4kX6x6TMPXpQSJJGyk0ywVUxDdDFzz+BA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-04T19:34:47.700551Z"},"content_sha256":"70a8008951322c0b4b660dd34bbf397d4d51eafaf1d12e838c15a3fee8525fa9","schema_version":"1.0","event_id":"sha256:70a8008951322c0b4b660dd34bbf397d4d51eafaf1d12e838c15a3fee8525fa9"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/IST3XHVNJSADRSBL5L6JRTVLNT/bundle.json","state_url":"https://pith.science/pith/IST3XHVNJSADRSBL5L6JRTVLNT/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/IST3XHVNJSADRSBL5L6JRTVLNT/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":"2026-06-04T19:34:47Z","links":{"resolver":"https://pith.science/pith/IST3XHVNJSADRSBL5L6JRTVLNT","bundle":"https://pith.science/pith/IST3XHVNJSADRSBL5L6JRTVLNT/bundle.json","state":"https://pith.science/pith/IST3XHVNJSADRSBL5L6JRTVLNT/state.json","well_known_bundle":"https://pith.science/.well-known/pith/IST3XHVNJSADRSBL5L6JRTVLNT/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:IST3XHVNJSADRSBL5L6JRTVLNT","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":"a5fa7e08c14cfe870725a2b0ab026cc82993bc6daed0af6062dcbe35be363a91","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-04-06T17:58:09Z","title_canon_sha256":"2da0c9d0dd1bb0ca8ff88374f47294a26d976e55871dbc5d18aa2d9f2cae193e"},"schema_version":"1.0","source":{"id":"2304.03277","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2304.03277","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"arxiv_version","alias_value":"2304.03277v1","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2304.03277","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"pith_short_12","alias_value":"IST3XHVNJSAD","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"IST3XHVNJSADRSBL","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"IST3XHVN","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:70a8008951322c0b4b660dd34bbf397d4d51eafaf1d12e838c15a3fee8525fa9","target":"graph","created_at":"2026-05-17T23:39:22Z","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":"Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That performance gains on the chosen zero-shot tasks reflect genuine improvement in instruction following rather than artifacts of GPT-4's data generation process or evaluation choices."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"GPT-4-generated instruction data produces superior zero-shot performance in finetuned LLaMA models versus prior state-of-the-art data."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"GPT-4 generated instruction data enables LLaMA models to reach higher zero-shot performance on new tasks than earlier synthetic datasets."}],"snapshot_sha256":"3e5864af4ec74bde1b4fb3b6168781c91bd2d762b48e55b99d23d6306483114d"},"formal_canon":{"evidence_count":1,"snapshot_sha256":"c26f16a8dcc7d1eb015e86884555556a48859d117881c5f57bf53f828be0fea0"},"paper":{"abstract_excerpt":"Prior work has shown that finetuning large language models (LLMs) using machine-generated instruction-following data enables such models to achieve remarkable zero-shot capabilities on new tasks, and no human-written instructions are needed. In this paper, we present the first attempt to use GPT-4 to generate instruction-following data for LLM finetuning. Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by pr","authors_text":"Baolin Peng, Chunyuan Li, Jianfeng Gao, Michel Galley, Pengcheng He","cross_cats":["cs.AI"],"headline":"GPT-4 generated instruction data enables LLaMA models to reach higher zero-shot performance on new tasks than earlier synthetic datasets.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-04-06T17:58:09Z","title":"Instruction Tuning with GPT-4"},"references":{"count":18,"internal_anchors":11,"resolved_work":18,"sample":[{"cited_arxiv_id":"2112.00861","doi":"","is_internal_anchor":true,"ref_index":1,"title":"A General Language Assistant as a Laboratory for Alignment","work_id":"a43f9ea0-01be-47d5-b8ee-a1a9f73381c5","year":null},{"cited_arxiv_id":"","doi":"10.5281/zenodo.7733589","is_internal_anchor":false,"ref_index":2,"title":"Yejin Bang, Samuel Cahyawijaya, Nayeon Lee, Wenliang Dai, Dan Su, Bryan Wilie, Holy Lovenia, Ziwei Ji, Tiezheng Yu, Willy Chung, et al","work_id":"6f218053-cca5-4fde-92aa-730589931f0c","year":null},{"cited_arxiv_id":"2212.08073","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Constitutional AI: Harmlessness from AI Feedback","work_id":"faaaa4e0-2676-4fac-a0b4-99aef10d2095","year":null},{"cited_arxiv_id":"","doi":"10.5281/zenodo.5297715","is_internal_anchor":false,"ref_index":4,"title":"GPT-Neo: Large Scale Autoregressive Language Modeling with Mesh-Tensorflow , March 2021","work_id":"6c7f8a44-6f52-448c-b819-5ba82a7bbc59","year":1901},{"cited_arxiv_id":"2210.11416","doi":"","is_internal_anchor":true,"ref_index":5,"title":"Scaling Instruction-Finetuned Language Models","work_id":"8405abb1-7558-4fdf-af24-f4c52fa77a06","year":null}],"snapshot_sha256":"086a76bdf9860cd1a85670949cbde7831e84f7f71ab1d678f23db21772a9c2ed"},"source":{"id":"2304.03277","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T16:57:59.957977Z","id":"7170c55c-42cb-42e8-b30e-0c41f8928e3d","model_set":{"reader":"grok-4.3"},"one_line_summary":"GPT-4-generated instruction data produces superior zero-shot performance in finetuned LLaMA models versus prior state-of-the-art data.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"GPT-4 generated instruction data enables LLaMA models to reach higher zero-shot performance on new tasks than earlier synthetic datasets.","strongest_claim":"Our early experiments on instruction-tuned LLaMA models show that the 52K English and Chinese instruction-following data generated by GPT-4 leads to superior zero-shot performance on new tasks to the instruction-following data generated by previous state-of-the-art models.","weakest_assumption":"That performance gains on the chosen zero-shot tasks reflect genuine improvement in instruction following rather than artifacts of GPT-4's data generation process or evaluation choices."}},"verdict_id":"7170c55c-42cb-42e8-b30e-0c41f8928e3d"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:a32cc948bf1a828856f5d88ce7603244c9939bc0ed57a9823689f514a558a216","target":"record","created_at":"2026-05-17T23:39:22Z","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":"a5fa7e08c14cfe870725a2b0ab026cc82993bc6daed0af6062dcbe35be363a91","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2023-04-06T17:58:09Z","title_canon_sha256":"2da0c9d0dd1bb0ca8ff88374f47294a26d976e55871dbc5d18aa2d9f2cae193e"},"schema_version":"1.0","source":{"id":"2304.03277","kind":"arxiv","version":1}},"canonical_sha256":"44a7bb9ead4c8038c82beafc98ceab6ce87f1de79c8e0b9e50615d9cc1fee7af","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"44a7bb9ead4c8038c82beafc98ceab6ce87f1de79c8e0b9e50615d9cc1fee7af","first_computed_at":"2026-05-17T23:39:22.359932Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:22.359932Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"S+UED9yYs0+zmAMcnpwQloKfVlHgac9qhW074tDP0h0a3aGJj/IFy9dMI9WPbE3j1ApZ4w4ww/V7Q2v3N0DzAA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:22.360701Z","signed_message":"canonical_sha256_bytes"},"source_id":"2304.03277","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:a32cc948bf1a828856f5d88ce7603244c9939bc0ed57a9823689f514a558a216","sha256:70a8008951322c0b4b660dd34bbf397d4d51eafaf1d12e838c15a3fee8525fa9"],"state_sha256":"e9f99c2acbde2d962e8de144c7543f6c9b52c37e3a8b51442c4db9f576816590"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"mDAIB3w31BU3O7PykwLcSXyTQ8hILEiUtWc81ly7UpOQbSNbprfn41ackvpoLW1+mfPm8BUW4IZaH/38rq+HDw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-04T19:34:47.703016Z","bundle_sha256":"f656efe322dd6b08eef163a412eac2c2e4313c4b4d393082ec3f8ad2dc4b8248"}}