{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:VV7D6X4NKA5XSLSHIYQRE5SR3I","short_pith_number":"pith:VV7D6X4N","canonical_record":{"source":{"id":"2309.05463","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T14:01:45Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"99f1dc3a00721b4964d2f8fcf43443912ce5445816a292edfbd4c7d6b695e336","abstract_canon_sha256":"f2258d654a2f0d6c28d086ba9fe59e7ffe77bc4cefb4f99f008ef8e38d28238b"},"schema_version":"1.0"},"canonical_sha256":"ad7e3f5f8d503b792e474621127651da1367db03a76710f51beb0b221c4806d7","source":{"kind":"arxiv","id":"2309.05463","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2309.05463","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"arxiv_version","alias_value":"2309.05463v1","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.05463","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"pith_short_12","alias_value":"VV7D6X4NKA5X","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"VV7D6X4NKA5XSLSH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"VV7D6X4N","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:VV7D6X4NKA5XSLSHIYQRE5SR3I","target":"record","payload":{"canonical_record":{"source":{"id":"2309.05463","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T14:01:45Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"99f1dc3a00721b4964d2f8fcf43443912ce5445816a292edfbd4c7d6b695e336","abstract_canon_sha256":"f2258d654a2f0d6c28d086ba9fe59e7ffe77bc4cefb4f99f008ef8e38d28238b"},"schema_version":"1.0"},"canonical_sha256":"ad7e3f5f8d503b792e474621127651da1367db03a76710f51beb0b221c4806d7","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:22.090289Z","signature_b64":"GnBaG4cRjpfsH+8ProSR9s7skzXY54SGWCmW9p4iIYjP2FuHBNHXBeAP2gdu+Bfnfi5kCfsbcC2Iq2GSgrRlDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ad7e3f5f8d503b792e474621127651da1367db03a76710f51beb0b221c4806d7","last_reissued_at":"2026-05-17T23:39:22.089627Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:22.089627Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2309.05463","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":"LgvF4ZuvCTSSlue+1We3hCghk7GrocSenIntycHkgSFW4DwOxHI5G3bGz64H7bRJfuEa9coZGmBtGYrhxYcuDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-07T10:27:27.540384Z"},"content_sha256":"33d930ed98ee8d61382c7cbc6181f33bd681a770704156d9aa07c9a4f2092a4c","schema_version":"1.0","event_id":"sha256:33d930ed98ee8d61382c7cbc6181f33bd681a770704156d9aa07c9a4f2092a4c"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:VV7D6X4NKA5XSLSHIYQRE5SR3I","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Textbooks Are All You Need II: phi-1.5 technical report","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A 1.3 billion parameter model trained on synthetic textbooks matches models five times larger on reasoning tasks.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Allie Del Giorno, Ronen Eldan, S\\'ebastien Bubeck, Suriya Gunasekar, Yin Tat Lee, Yuanzhi Li","submitted_at":"2023-09-11T14:01:45Z","abstract_excerpt":"We continue the investigation into the power of smaller Transformer-based language models as initiated by \\textbf{TinyStories} -- a 10 million parameter model that can produce coherent English -- and the follow-up work on \\textbf{phi-1}, a 1.3 billion parameter model with Python coding performance close to the state-of-the-art. The latter work proposed to use existing Large Language Models (LLMs) to generate ``textbook quality\" data as a way to enhance the learning process compared to traditional web data. We follow the ``Textbooks Are All You Need\" approach, focusing this time on common sense"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"phi-1.5 ... with performance on natural language tasks comparable to models 5x larger, and surpassing most non-frontier LLMs on more complex reasoning tasks such as grade-school mathematics and basic coding.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That standard benchmarks for grade-school math and basic coding sufficiently measure general common sense reasoning without the model overfitting to patterns in the synthetic textbook data.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"phi-1.5 is a 1.3B parameter model trained on synthetic textbook data that matches the reasoning performance of models five times larger on natural language, math, and basic coding tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A 1.3 billion parameter model trained on synthetic textbooks matches models five times larger on reasoning tasks.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"bd49c7239b4b89d801f193cb561cadc90abf68ae81b2ff13704a29d11ade9841"},"source":{"id":"2309.05463","kind":"arxiv","version":1},"verdict":{"id":"29c3349e-ead8-483e-b856-738cf676389a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:13:13.944635Z","strongest_claim":"phi-1.5 ... with performance on natural language tasks comparable to models 5x larger, and surpassing most non-frontier LLMs on more complex reasoning tasks such as grade-school mathematics and basic coding.","one_line_summary":"phi-1.5 is a 1.3B parameter model trained on synthetic textbook data that matches the reasoning performance of models five times larger on natural language, math, and basic coding tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That standard benchmarks for grade-school math and basic coding sufficiently measure general common sense reasoning without the model overfitting to patterns in the synthetic textbook data.","pith_extraction_headline":"A 1.3 billion parameter model trained on synthetic textbooks matches models five times larger on reasoning tasks."},"references":{"count":24,"sample":[{"doi":"","year":null,"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","ref_index":1,"cited_arxiv_id":"2108.07732","is_internal_anchor":true},{"doi":"","year":null,"title":"Identify, align, and integrate: Matching knowledge graphs to commonsense reasoning tasks","work_id":"9f1ee709-7c10-4e2e-a368-5342f156dbfa","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Sparks of Artificial General Intelligence: Early experiments with GPT-4","work_id":"a23cfe92-7f7c-424b-98d4-b386a83002fb","ref_index":3,"cited_arxiv_id":"2303.12712","is_internal_anchor":true},{"doi":"","year":2021,"title":"On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency , pages 610–623","work_id":"c49bc9e9-444d-46e0-bdec-fc27a248705a","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1911,"title":"PIQA: Reasoning about Physical Commonsense in Natural Language","work_id":"0d865a62-6376-4606-8d3a-eeb3b6e9ba6d","ref_index":5,"cited_arxiv_id":"1911.11641","is_internal_anchor":true}],"resolved_work":24,"snapshot_sha256":"baa61333aed4ee2a2ffbf29b8d56e89cb69cf4cf24f0ae4bb428cc66c971777d","internal_anchors":14},"formal_canon":{"evidence_count":2,"snapshot_sha256":"5e6edb4a9c5139f8230b9008f46ba426ec5f32b7731d9a02315bf3ad3947f19c"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"29c3349e-ead8-483e-b856-738cf676389a"},"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":"z8Gk5bHHHwoQJAiowMOJI/UyggGKFvddei+odLOirqxbTEYAnjG4OnuUJ7ATQmJce8R8gDqaZ7igsOC5nNJTDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-07T10:27:27.541274Z"},"content_sha256":"0e52fbcb3b6f62e07acb748f7ace481c778aad4ce949102cd145c21ffd598705","schema_version":"1.0","event_id":"sha256:0e52fbcb3b6f62e07acb748f7ace481c778aad4ce949102cd145c21ffd598705"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/bundle.json","state_url":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/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":"2026-06-07T10:27:27Z","links":{"resolver":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I","bundle":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/bundle.json","state":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/state.json","well_known_bundle":"https://pith.science/.well-known/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:VV7D6X4NKA5XSLSHIYQRE5SR3I","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":"f2258d654a2f0d6c28d086ba9fe59e7ffe77bc4cefb4f99f008ef8e38d28238b","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T14:01:45Z","title_canon_sha256":"99f1dc3a00721b4964d2f8fcf43443912ce5445816a292edfbd4c7d6b695e336"},"schema_version":"1.0","source":{"id":"2309.05463","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2309.05463","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"arxiv_version","alias_value":"2309.05463v1","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.05463","created_at":"2026-05-17T23:39:22Z"},{"alias_kind":"pith_short_12","alias_value":"VV7D6X4NKA5X","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"VV7D6X4NKA5XSLSH","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"VV7D6X4N","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:0e52fbcb3b6f62e07acb748f7ace481c778aad4ce949102cd145c21ffd598705","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":"phi-1.5 ... with performance on natural language tasks comparable to models 5x larger, and surpassing most non-frontier LLMs on more complex reasoning tasks such as grade-school mathematics and basic coding."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That standard benchmarks for grade-school math and basic coding sufficiently measure general common sense reasoning without the model overfitting to patterns in the synthetic textbook data."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"phi-1.5 is a 1.3B parameter model trained on synthetic textbook data that matches the reasoning performance of models five times larger on natural language, math, and basic coding tasks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"A 1.3 billion parameter model trained on synthetic textbooks matches models five times larger on reasoning tasks."}],"snapshot_sha256":"bd49c7239b4b89d801f193cb561cadc90abf68ae81b2ff13704a29d11ade9841"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"5e6edb4a9c5139f8230b9008f46ba426ec5f32b7731d9a02315bf3ad3947f19c"},"paper":{"abstract_excerpt":"We continue the investigation into the power of smaller Transformer-based language models as initiated by \\textbf{TinyStories} -- a 10 million parameter model that can produce coherent English -- and the follow-up work on \\textbf{phi-1}, a 1.3 billion parameter model with Python coding performance close to the state-of-the-art. The latter work proposed to use existing Large Language Models (LLMs) to generate ``textbook quality\" data as a way to enhance the learning process compared to traditional web data. We follow the ``Textbooks Are All You Need\" approach, focusing this time on common sense","authors_text":"Allie Del Giorno, Ronen Eldan, S\\'ebastien Bubeck, Suriya Gunasekar, Yin Tat Lee, Yuanzhi Li","cross_cats":["cs.AI"],"headline":"A 1.3 billion parameter model trained on synthetic textbooks matches models five times larger on reasoning tasks.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T14:01:45Z","title":"Textbooks Are All You Need II: phi-1.5 technical report"},"references":{"count":24,"internal_anchors":14,"resolved_work":24,"sample":[{"cited_arxiv_id":"2108.07732","doi":"","is_internal_anchor":true,"ref_index":1,"title":"Program Synthesis with Large Language Models","work_id":"fd241a05-03b9-4de2-9588-9d77ce176125","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Identify, align, and integrate: Matching knowledge graphs to commonsense reasoning tasks","work_id":"9f1ee709-7c10-4e2e-a368-5342f156dbfa","year":null},{"cited_arxiv_id":"2303.12712","doi":"","is_internal_anchor":true,"ref_index":3,"title":"Sparks of Artificial General Intelligence: Early experiments with GPT-4","work_id":"a23cfe92-7f7c-424b-98d4-b386a83002fb","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency , pages 610–623","work_id":"c49bc9e9-444d-46e0-bdec-fc27a248705a","year":2021},{"cited_arxiv_id":"1911.11641","doi":"","is_internal_anchor":true,"ref_index":5,"title":"PIQA: Reasoning about Physical Commonsense in Natural Language","work_id":"0d865a62-6376-4606-8d3a-eeb3b6e9ba6d","year":1911}],"snapshot_sha256":"baa61333aed4ee2a2ffbf29b8d56e89cb69cf4cf24f0ae4bb428cc66c971777d"},"source":{"id":"2309.05463","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:13:13.944635Z","id":"29c3349e-ead8-483e-b856-738cf676389a","model_set":{"reader":"grok-4.3"},"one_line_summary":"phi-1.5 is a 1.3B parameter model trained on synthetic textbook data that matches the reasoning performance of models five times larger on natural language, math, and basic coding tasks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"A 1.3 billion parameter model trained on synthetic textbooks matches models five times larger on reasoning tasks.","strongest_claim":"phi-1.5 ... with performance on natural language tasks comparable to models 5x larger, and surpassing most non-frontier LLMs on more complex reasoning tasks such as grade-school mathematics and basic coding.","weakest_assumption":"That standard benchmarks for grade-school math and basic coding sufficiently measure general common sense reasoning without the model overfitting to patterns in the synthetic textbook data."}},"verdict_id":"29c3349e-ead8-483e-b856-738cf676389a"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:33d930ed98ee8d61382c7cbc6181f33bd681a770704156d9aa07c9a4f2092a4c","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":"f2258d654a2f0d6c28d086ba9fe59e7ffe77bc4cefb4f99f008ef8e38d28238b","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T14:01:45Z","title_canon_sha256":"99f1dc3a00721b4964d2f8fcf43443912ce5445816a292edfbd4c7d6b695e336"},"schema_version":"1.0","source":{"id":"2309.05463","kind":"arxiv","version":1}},"canonical_sha256":"ad7e3f5f8d503b792e474621127651da1367db03a76710f51beb0b221c4806d7","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ad7e3f5f8d503b792e474621127651da1367db03a76710f51beb0b221c4806d7","first_computed_at":"2026-05-17T23:39:22.089627Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:22.089627Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"GnBaG4cRjpfsH+8ProSR9s7skzXY54SGWCmW9p4iIYjP2FuHBNHXBeAP2gdu+Bfnfi5kCfsbcC2Iq2GSgrRlDw==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:22.090289Z","signed_message":"canonical_sha256_bytes"},"source_id":"2309.05463","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:33d930ed98ee8d61382c7cbc6181f33bd681a770704156d9aa07c9a4f2092a4c","sha256:0e52fbcb3b6f62e07acb748f7ace481c778aad4ce949102cd145c21ffd598705"],"state_sha256":"bda9183628778498eba7773c40632a899e32f3e0fa0fe4916b7c5d13891728b1"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"RAw2gDfrAkmg9I/iVergkUH61emsgWWIKGxZk4Se9XhEH8vVnmA1tg5wkj3NPV58HlU6Vaay/BXPKggWeK2vAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-07T10:27:27.550683Z","bundle_sha256":"604a1a89eb4910a63819321f80a6f9a7bbecd93f67b1d6bdd163780fb151f2c9"}}