{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:VV7D6X4NKA5XSLSHIYQRE5SR3I","short_pith_number":"pith:VV7D6X4N","schema_version":"1.0","canonical_sha256":"ad7e3f5f8d503b792e474621127651da1367db03a76710f51beb0b221c4806d7","source":{"kind":"arxiv","id":"2309.05463","version":1},"attestation_state":"computed","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"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":true,"formal_links_present":true},"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"},"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"},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2309.05463","created_at":"2026-05-17T23:39:22.089738+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.05463v1","created_at":"2026-05-17T23:39:22.089738+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.05463","created_at":"2026-05-17T23:39:22.089738+00:00"},{"alias_kind":"pith_short_12","alias_value":"VV7D6X4NKA5X","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"VV7D6X4NKA5XSLSH","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"VV7D6X4N","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":40,"internal_anchor_count":40,"sample":[{"citing_arxiv_id":"2406.11354","citing_title":"Preserving Knowledge in Large Language Model with Model-Agnostic Self-Decompression","ref_index":32,"is_internal_anchor":true},{"citing_arxiv_id":"2410.20791","citing_title":"From Cool Demos to Production-Ready FMware: Core Challenges and a Technology Roadmap","ref_index":70,"is_internal_anchor":true},{"citing_arxiv_id":"2410.23728","citing_title":"GigaCheck: Detecting LLM-generated Content via Object-Centric Span Localization","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2501.01046","citing_title":"SEDD: Scalable and Efficient Dataset Deduplication with GPUs","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.22579","citing_title":"Beyond Temperature: Hyperfitting as a Late-Stage Geometric Expansion","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21539","citing_title":"DualOptim+: Bridging Shared and Decoupled Optimizer States for Better Machine Unlearning in Large Language Models","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2602.08167","citing_title":"Self-Supervised Bootstrapping of Action-Predictive Embodied Reasoning","ref_index":57,"is_internal_anchor":true},{"citing_arxiv_id":"2412.13663","citing_title":"Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference","ref_index":163,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16776","citing_title":"Distinguishable Deletion: Unifying Knowledge Erasure and Refusal for Large Language Model Unlearning","ref_index":47,"is_internal_anchor":true},{"citing_arxiv_id":"2506.03535","citing_title":"Across Programming Language Silos: A Study on Cross-Lingual Retrieval-augmented Code Generation","ref_index":51,"is_internal_anchor":true},{"citing_arxiv_id":"2506.06226","citing_title":"No Data? No Problem: Synthesizing Security Graphs for Better Intrusion Detection","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2402.13116","citing_title":"A Survey on Knowledge Distillation of Large Language Models","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2406.11794","citing_title":"DataComp-LM: In search of the next generation of training sets for language models","ref_index":107,"is_internal_anchor":true},{"citing_arxiv_id":"2511.23332","citing_title":"UniGeoSeg: Towards Unified Open-World Segmentation for Geospatial Scenes","ref_index":31,"is_internal_anchor":true},{"citing_arxiv_id":"2505.16933","citing_title":"LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2601.02627","citing_title":"Improved Evidence Extraction and Metrics for Document Inconsistency Detection with LLMs","ref_index":3,"is_internal_anchor":true},{"citing_arxiv_id":"2401.06121","citing_title":"TOFU: A Task of Fictitious Unlearning for LLMs","ref_index":19,"is_internal_anchor":true},{"citing_arxiv_id":"2311.16867","citing_title":"The Falcon Series of Open Language Models","ref_index":230,"is_internal_anchor":true},{"citing_arxiv_id":"2403.07691","citing_title":"ORPO: Monolithic Preference Optimization without Reference Model","ref_index":33,"is_internal_anchor":true},{"citing_arxiv_id":"2603.00655","citing_title":"Mema: Memory-Augmented Adapter for Enhanced Vision-Language Understanding","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2604.02714","citing_title":"ExploreVLA: Dense World Modeling and Exploration for End-to-End Autonomous Driving","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2406.00515","citing_title":"A Survey on Large Language Models for Code Generation","ref_index":153,"is_internal_anchor":true},{"citing_arxiv_id":"2502.02737","citing_title":"SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model","ref_index":196,"is_internal_anchor":true},{"citing_arxiv_id":"2311.05232","citing_title":"A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions","ref_index":184,"is_internal_anchor":true},{"citing_arxiv_id":"2604.24819","citing_title":"Programming with Data: Test-Driven Data Engineering for Self-Improving LLMs from Raw Corpora","ref_index":27,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I","json":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I.json","graph_json":"https://pith.science/api/pith-number/VV7D6X4NKA5XSLSHIYQRE5SR3I/graph.json","events_json":"https://pith.science/api/pith-number/VV7D6X4NKA5XSLSHIYQRE5SR3I/events.json","paper":"https://pith.science/paper/VV7D6X4N"},"agent_actions":{"view_html":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I","download_json":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I.json","view_paper":"https://pith.science/paper/VV7D6X4N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.05463&json=true","fetch_graph":"https://pith.science/api/pith-number/VV7D6X4NKA5XSLSHIYQRE5SR3I/graph.json","fetch_events":"https://pith.science/api/pith-number/VV7D6X4NKA5XSLSHIYQRE5SR3I/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/action/storage_attestation","attest_author":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/action/author_attestation","sign_citation":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/action/citation_signature","submit_replication":"https://pith.science/pith/VV7D6X4NKA5XSLSHIYQRE5SR3I/action/replication_record"}},"created_at":"2026-05-17T23:39:22.089738+00:00","updated_at":"2026-05-17T23:39:22.089738+00:00"}