{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:4THWGCRDHG5VQM6PZLGWHG25GG","short_pith_number":"pith:4THWGCRD","schema_version":"1.0","canonical_sha256":"e4cf630a2339bb5833cfcacd639b5d319a65a46d298406effc0cb99595cd5db8","source":{"kind":"arxiv","id":"2607.04733","version":1},"attestation_state":"computed","paper":{"title":"LP-SFT: Local-Preserving Supervised Fine-Tuning via Multimodal Entropy Structure","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Baolong Bi, Jingyuan Zhang, Shuo Lu, Yueyang Wang","submitted_at":"2026-07-06T07:14:22Z","abstract_excerpt":"Supervised fine-tuning (SFT) is the standard approach for adapting pretrained language models to downstream domains, yet it often improves target-domain behavior at the cost of degrading pre-existing capabilities. Standard cross-entropy fine-tuning promotes only the observed label token and leaves unconstrained how probability mass is redistributed over other plausible alternatives, potentially distorting the rich local preference structure learned during pretraining. We first analyze next-token predictions using Shannon and Renyi entropies, revealing that pretrained models exhibit a regular m"},"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":"2607.04733","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-07-06T07:14:22Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ff78918336640ddf3e5ba5974236430cd44b28c38e8c843dd70806d77a2753a9","abstract_canon_sha256":"1a3d08152379be578983a506fc8ef93bdf25cf56377da8012c69eda549939628"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-07T02:20:00.553307Z","signature_b64":"pxvpUR8uwc+53rJ/VDhIeqQhcXo2o7IvWkWney44J9gZYfLC730/cAriEM5tGVRht7xCeYvAEjJLTJaYQLZnCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e4cf630a2339bb5833cfcacd639b5d319a65a46d298406effc0cb99595cd5db8","last_reissued_at":"2026-07-07T02:20:00.552495Z","signature_status":"signed_v1","first_computed_at":"2026-07-07T02:20:00.552495Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LP-SFT: Local-Preserving Supervised Fine-Tuning via Multimodal Entropy Structure","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Baolong Bi, Jingyuan Zhang, Shuo Lu, Yueyang Wang","submitted_at":"2026-07-06T07:14:22Z","abstract_excerpt":"Supervised fine-tuning (SFT) is the standard approach for adapting pretrained language models to downstream domains, yet it often improves target-domain behavior at the cost of degrading pre-existing capabilities. Standard cross-entropy fine-tuning promotes only the observed label token and leaves unconstrained how probability mass is redistributed over other plausible alternatives, potentially distorting the rich local preference structure learned during pretraining. We first analyze next-token predictions using Shannon and Renyi entropies, revealing that pretrained models exhibit a regular m"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.04733","kind":"arxiv","version":1},"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/2607.04733/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":"2607.04733","created_at":"2026-07-07T02:20:00.552640+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.04733v1","created_at":"2026-07-07T02:20:00.552640+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.04733","created_at":"2026-07-07T02:20:00.552640+00:00"},{"alias_kind":"pith_short_12","alias_value":"4THWGCRDHG5V","created_at":"2026-07-07T02:20:00.552640+00:00"},{"alias_kind":"pith_short_16","alias_value":"4THWGCRDHG5VQM6P","created_at":"2026-07-07T02:20:00.552640+00:00"},{"alias_kind":"pith_short_8","alias_value":"4THWGCRD","created_at":"2026-07-07T02:20:00.552640+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG","json":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG.json","graph_json":"https://pith.science/api/pith-number/4THWGCRDHG5VQM6PZLGWHG25GG/graph.json","events_json":"https://pith.science/api/pith-number/4THWGCRDHG5VQM6PZLGWHG25GG/events.json","paper":"https://pith.science/paper/4THWGCRD"},"agent_actions":{"view_html":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG","download_json":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG.json","view_paper":"https://pith.science/paper/4THWGCRD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.04733&json=true","fetch_graph":"https://pith.science/api/pith-number/4THWGCRDHG5VQM6PZLGWHG25GG/graph.json","fetch_events":"https://pith.science/api/pith-number/4THWGCRDHG5VQM6PZLGWHG25GG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG/action/storage_attestation","attest_author":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG/action/author_attestation","sign_citation":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG/action/citation_signature","submit_replication":"https://pith.science/pith/4THWGCRDHG5VQM6PZLGWHG25GG/action/replication_record"}},"created_at":"2026-07-07T02:20:00.552640+00:00","updated_at":"2026-07-07T02:20:00.552640+00:00"}