{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:NADBFOOLYGJNGQAEI2IQGK5OVG","short_pith_number":"pith:NADBFOOL","schema_version":"1.0","canonical_sha256":"680612b9cbc192d340044691032baea991a71dd336d9ad2210dce4f4916f574c","source":{"kind":"arxiv","id":"2505.12043","version":2},"attestation_state":"computed","paper":{"title":"MoL for LLMs: Dual-Loss Optimization to Enhance Domain Expertise While Preserving General Capabilities","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jingxue Chen, Qianchun Lu, Qingkun Tang, Siyuan Fang","submitted_at":"2025-05-17T15:12:47Z","abstract_excerpt":"Although large language models (LLMs) perform well in general tasks, domain-specific applications suffer from hallucinations and accuracy limitations. Continual Pre-Training (CPT) approaches encounter two key issues: (1) domain-biased data degrades general language skills, and (2) improper corpus-mixture ratios limit effective adaptation. To address these, we propose a novel framework, Mixture of Losses (MoL), which decouples optimization objectives for domain-specific and general corpora. Specifically, cross-entropy (CE) loss is applied to domain-corpus to ensure knowledge acquisition, while "},"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":"2505.12043","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-17T15:12:47Z","cross_cats_sorted":[],"title_canon_sha256":"2f24bfa8859bf9ca1ba1da9d02922bfe6fb99de5582c2c075d732c8a3a222e8e","abstract_canon_sha256":"9c13e3561321c32109d949b0b59f29adf95ef4e19e4d8ac0bb4d89d56379209a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:45.160247Z","signature_b64":"qmXq6+EWur6F8Vnjz+Qqlldv43L1p+0/cCyxYDEEmgiRGJmPQcnCwtMfyn0v2B8IjjELlR7wUSzpet2Y3YdcCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"680612b9cbc192d340044691032baea991a71dd336d9ad2210dce4f4916f574c","last_reissued_at":"2026-07-05T11:05:45.159783Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:45.159783Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MoL for LLMs: Dual-Loss Optimization to Enhance Domain Expertise While Preserving General Capabilities","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jingxue Chen, Qianchun Lu, Qingkun Tang, Siyuan Fang","submitted_at":"2025-05-17T15:12:47Z","abstract_excerpt":"Although large language models (LLMs) perform well in general tasks, domain-specific applications suffer from hallucinations and accuracy limitations. Continual Pre-Training (CPT) approaches encounter two key issues: (1) domain-biased data degrades general language skills, and (2) improper corpus-mixture ratios limit effective adaptation. To address these, we propose a novel framework, Mixture of Losses (MoL), which decouples optimization objectives for domain-specific and general corpora. Specifically, cross-entropy (CE) loss is applied to domain-corpus to ensure knowledge acquisition, while "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.12043","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/2505.12043/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":"2505.12043","created_at":"2026-07-05T11:05:45.159834+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.12043v2","created_at":"2026-07-05T11:05:45.159834+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.12043","created_at":"2026-07-05T11:05:45.159834+00:00"},{"alias_kind":"pith_short_12","alias_value":"NADBFOOLYGJN","created_at":"2026-07-05T11:05:45.159834+00:00"},{"alias_kind":"pith_short_16","alias_value":"NADBFOOLYGJNGQAE","created_at":"2026-07-05T11:05:45.159834+00:00"},{"alias_kind":"pith_short_8","alias_value":"NADBFOOL","created_at":"2026-07-05T11:05:45.159834+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.07753","citing_title":"Symbiotic-MoE: Unlocking the Synergy between Generation and Understanding","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG","json":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG.json","graph_json":"https://pith.science/api/pith-number/NADBFOOLYGJNGQAEI2IQGK5OVG/graph.json","events_json":"https://pith.science/api/pith-number/NADBFOOLYGJNGQAEI2IQGK5OVG/events.json","paper":"https://pith.science/paper/NADBFOOL"},"agent_actions":{"view_html":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG","download_json":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG.json","view_paper":"https://pith.science/paper/NADBFOOL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.12043&json=true","fetch_graph":"https://pith.science/api/pith-number/NADBFOOLYGJNGQAEI2IQGK5OVG/graph.json","fetch_events":"https://pith.science/api/pith-number/NADBFOOLYGJNGQAEI2IQGK5OVG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG/action/storage_attestation","attest_author":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG/action/author_attestation","sign_citation":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG/action/citation_signature","submit_replication":"https://pith.science/pith/NADBFOOLYGJNGQAEI2IQGK5OVG/action/replication_record"}},"created_at":"2026-07-05T11:05:45.159834+00:00","updated_at":"2026-07-05T11:05:45.159834+00:00"}