{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:POXFCKP4TR7IOA6MILVBT5E3DL","short_pith_number":"pith:POXFCKP4","schema_version":"1.0","canonical_sha256":"7bae5129fc9c7e8703cc42ea19f49b1ac75b80ab4312b207dea607f631759975","source":{"kind":"arxiv","id":"2310.05492","version":4},"attestation_state":"computed","paper":{"title":"How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chang Zhou, Chengpeng Li, Dayiheng Liu, Guanting Dong, Hongyi Yuan, Jingren Zhou, Keming Lu, Mingfeng Xue, Wei Wang, Zheng Yuan","submitted_at":"2023-10-09T07:56:16Z","abstract_excerpt":"Large language models (LLMs) with enormous pre-training tokens and parameters emerge diverse abilities, including math reasoning, code generation, and instruction following. These abilities are further enhanced by supervised fine-tuning (SFT). While the open-source community has explored ad-hoc SFT for enhancing individual capabilities, proprietary LLMs exhibit versatility across various skills. Therefore, understanding the facilitation of multiple abilities via SFT is paramount. In this study, we specifically focuses on the interplay of data composition between mathematical reasoning, code ge"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2310.05492","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-10-09T07:56:16Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"0fb75604f26716dd0c8fbe6a1c7437afc6a39260c08a54be57719d8de0f63052","abstract_canon_sha256":"57efea4453f8f163dba40b6902450e8cc1e9488e0f28af0179fad410ea5b457f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:34.156910Z","signature_b64":"5c7OB6SLBnNxvngBqZKGKktQxyJvC78Sqzgw6DHWe2NIMR21fq0adUND1VcB/Bo0dVlwXRlpOG32V5TS98ukBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7bae5129fc9c7e8703cc42ea19f49b1ac75b80ab4312b207dea607f631759975","last_reissued_at":"2026-07-05T08:28:34.156336Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:34.156336Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chang Zhou, Chengpeng Li, Dayiheng Liu, Guanting Dong, Hongyi Yuan, Jingren Zhou, Keming Lu, Mingfeng Xue, Wei Wang, Zheng Yuan","submitted_at":"2023-10-09T07:56:16Z","abstract_excerpt":"Large language models (LLMs) with enormous pre-training tokens and parameters emerge diverse abilities, including math reasoning, code generation, and instruction following. These abilities are further enhanced by supervised fine-tuning (SFT). While the open-source community has explored ad-hoc SFT for enhancing individual capabilities, proprietary LLMs exhibit versatility across various skills. Therefore, understanding the facilitation of multiple abilities via SFT is paramount. In this study, we specifically focuses on the interplay of data composition between mathematical reasoning, code ge"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.05492","kind":"arxiv","version":4},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2310.05492/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2310.05492","created_at":"2026-07-05T08:28:34.156412+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.05492v4","created_at":"2026-07-05T08:28:34.156412+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.05492","created_at":"2026-07-05T08:28:34.156412+00:00"},{"alias_kind":"pith_short_12","alias_value":"POXFCKP4TR7I","created_at":"2026-07-05T08:28:34.156412+00:00"},{"alias_kind":"pith_short_16","alias_value":"POXFCKP4TR7IOA6M","created_at":"2026-07-05T08:28:34.156412+00:00"},{"alias_kind":"pith_short_8","alias_value":"POXFCKP4","created_at":"2026-07-05T08:28:34.156412+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00604","citing_title":"Vehicle Routing Problem Meets Large Language Models: An Overview and Perspectives","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2506.01247","citing_title":"Beyond Interpretability: When, Why, and How Sparse Autoencoders Enable Label-Free Visual Steering","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2502.10248","citing_title":"Step-Video-T2V Technical Report: The Practice, Challenges, and Future of Video Foundation Model","ref_index":132,"is_internal_anchor":false},{"citing_arxiv_id":"2510.09689","citing_title":"When Search Goes Wrong: Red-Teaming Web-Augmented Large Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2403.07691","citing_title":"ORPO: Monolithic Preference Optimization without Reference Model","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2507.21046","citing_title":"A Survey of Self-Evolving Agents: What, When, How, and Where to Evolve on the Path to Artificial Super Intelligence","ref_index":135,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13225","citing_title":"Mix, Don't Tune: Bilingual Pre-Training Outperforms Hyperparameter Search in Data-Constrained Settings","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09533","citing_title":"Assessment of RAG and Fine-Tuning for Industrial Question-Answering-Applications","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01123","citing_title":"PERSA: Reinforcement Learning for Professor-Style Personalized Feedback with LLMs","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04066","citing_title":"Adapt to Thrive! Adaptive Power-Mean Policy Optimization for Improved LLM Reasoning","ref_index":224,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04065","citing_title":"Free Energy-Driven Reinforcement Learning with Adaptive Advantage Shaping for Unsupervised Reasoning in LLMs","ref_index":239,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17184","citing_title":"SynthFix: Adaptive Neuro-Symbolic Code Vulnerability Repair","ref_index":83,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL","json":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL.json","graph_json":"https://pith.science/api/pith-number/POXFCKP4TR7IOA6MILVBT5E3DL/graph.json","events_json":"https://pith.science/api/pith-number/POXFCKP4TR7IOA6MILVBT5E3DL/events.json","paper":"https://pith.science/paper/POXFCKP4"},"agent_actions":{"view_html":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL","download_json":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL.json","view_paper":"https://pith.science/paper/POXFCKP4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.05492&json=true","fetch_graph":"https://pith.science/api/pith-number/POXFCKP4TR7IOA6MILVBT5E3DL/graph.json","fetch_events":"https://pith.science/api/pith-number/POXFCKP4TR7IOA6MILVBT5E3DL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL/action/storage_attestation","attest_author":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL/action/author_attestation","sign_citation":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL/action/citation_signature","submit_replication":"https://pith.science/pith/POXFCKP4TR7IOA6MILVBT5E3DL/action/replication_record"}},"created_at":"2026-07-05T08:28:34.156412+00:00","updated_at":"2026-07-05T08:28:34.156412+00:00"}