{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:R5CZKOTSW3XWYMZFNJAAJX2WNH","short_pith_number":"pith:R5CZKOTS","schema_version":"1.0","canonical_sha256":"8f45953a72b6ef6c33256a4004df5669cb1efc4c85bf3b4cf028baa8a633e52a","source":{"kind":"arxiv","id":"2502.11258","version":2},"attestation_state":"computed","paper":{"title":"Leveraging Conditional Mutual Information to Improve Large Language Model Fine-Tuning For Classification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"En-hui Yang, Thanushon Sivakaran","submitted_at":"2025-02-16T20:24:00Z","abstract_excerpt":"Although large language models (LLMs) have demonstrated remarkable capabilities in recent years, the potential of information theory (IT) to enhance LLM development remains underexplored. This paper introduces the information theoretic principle of Conditional Mutual Information (CMI) to LLM fine-tuning for classification tasks, exploring its promise in two main ways: minimizing CMI to improve a model's standalone performance and maximizing CMI to enhance knowledge distillation (KD) for more capable student models. To apply CMI in LLM fine-tuning, we adapt the recently proposed CMI-constrained"},"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":"2502.11258","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-16T20:24:00Z","cross_cats_sorted":[],"title_canon_sha256":"db79d4a1dbe4153ecbfa3d055b217aa6d64d0d9e2be6c39db0d5482545d40030","abstract_canon_sha256":"a7ad82a03d324faa09f5dda9a20d9e0ec60d31a6a3aed6b86b5d71d8a7349d8c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:55:54.619972Z","signature_b64":"8p4/lZ2+FDdltSPO7TpJzoVKzfyJLo1Gg8Evu43xNLoeWSUXTxPSnrZUmmRg6m5baiG3Zcmh4sa13vnW/qOXDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8f45953a72b6ef6c33256a4004df5669cb1efc4c85bf3b4cf028baa8a633e52a","last_reissued_at":"2026-07-05T10:55:54.619485Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:55:54.619485Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Leveraging Conditional Mutual Information to Improve Large Language Model Fine-Tuning For Classification","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"En-hui Yang, Thanushon Sivakaran","submitted_at":"2025-02-16T20:24:00Z","abstract_excerpt":"Although large language models (LLMs) have demonstrated remarkable capabilities in recent years, the potential of information theory (IT) to enhance LLM development remains underexplored. This paper introduces the information theoretic principle of Conditional Mutual Information (CMI) to LLM fine-tuning for classification tasks, exploring its promise in two main ways: minimizing CMI to improve a model's standalone performance and maximizing CMI to enhance knowledge distillation (KD) for more capable student models. To apply CMI in LLM fine-tuning, we adapt the recently proposed CMI-constrained"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.11258","kind":"arxiv","version":2},"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/2502.11258/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":"2502.11258","created_at":"2026-07-05T10:55:54.619541+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.11258v2","created_at":"2026-07-05T10:55:54.619541+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.11258","created_at":"2026-07-05T10:55:54.619541+00:00"},{"alias_kind":"pith_short_12","alias_value":"R5CZKOTSW3XW","created_at":"2026-07-05T10:55:54.619541+00:00"},{"alias_kind":"pith_short_16","alias_value":"R5CZKOTSW3XWYMZF","created_at":"2026-07-05T10:55:54.619541+00:00"},{"alias_kind":"pith_short_8","alias_value":"R5CZKOTS","created_at":"2026-07-05T10:55:54.619541+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.23790","citing_title":"Rethinking the Understanding Ability across LLMs through Mutual Information","ref_index":33,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH","json":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH.json","graph_json":"https://pith.science/api/pith-number/R5CZKOTSW3XWYMZFNJAAJX2WNH/graph.json","events_json":"https://pith.science/api/pith-number/R5CZKOTSW3XWYMZFNJAAJX2WNH/events.json","paper":"https://pith.science/paper/R5CZKOTS"},"agent_actions":{"view_html":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH","download_json":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH.json","view_paper":"https://pith.science/paper/R5CZKOTS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.11258&json=true","fetch_graph":"https://pith.science/api/pith-number/R5CZKOTSW3XWYMZFNJAAJX2WNH/graph.json","fetch_events":"https://pith.science/api/pith-number/R5CZKOTSW3XWYMZFNJAAJX2WNH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH/action/storage_attestation","attest_author":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH/action/author_attestation","sign_citation":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH/action/citation_signature","submit_replication":"https://pith.science/pith/R5CZKOTSW3XWYMZFNJAAJX2WNH/action/replication_record"}},"created_at":"2026-07-05T10:55:54.619541+00:00","updated_at":"2026-07-05T10:55:54.619541+00:00"}