{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:UTCXYALO7OUIRATAIH6MDJIFZU","short_pith_number":"pith:UTCXYALO","schema_version":"1.0","canonical_sha256":"a4c57c016efba888826041fcc1a505cd3a375bc09eccc43448e81a2912289b35","source":{"kind":"arxiv","id":"2505.13878","version":3},"attestation_state":"computed","paper":{"title":"InfiFPO: Implicit Model Fusion via Preference Optimization in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Fei Wu, Hongxia Yang, Qi Zhou, Yanggan Gu, Yiming Zhang, Yuanyi Wang, Zhaoyi Yan","submitted_at":"2025-05-20T03:32:37Z","abstract_excerpt":"Model fusion combines multiple Large Language Models (LLMs) with different strengths into a more powerful, integrated model through lightweight training methods. Existing works on model fusion focus primarily on supervised fine-tuning (SFT), leaving preference alignment (PA) --a critical phase for enhancing LLM performance--largely unexplored. The current few fusion methods on PA phase, like WRPO, simplify the process by utilizing only response outputs from source models while discarding their probability information. To address this limitation, we propose InfiFPO, a preference optimization me"},"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":"2505.13878","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-20T03:32:37Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"5fe2cd4904e2adb00ac64f1edc0b7a71f8ca5196eadf6de6cd08ec98a48ad5dc","abstract_canon_sha256":"660a860491d8a36a1d59c7def7f2bade8e683f79a5aab0f29b21bcfaafdd0a55"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T01:03:12.706481Z","signature_b64":"3vUHyzNxrKk2fnZ7u5eRZiVmubmqI54kAA1movhZVb9zRcsewwAIcMS2OvJ1C6NR0ZqYtkTgNTTAOuJrl7MbDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a4c57c016efba888826041fcc1a505cd3a375bc09eccc43448e81a2912289b35","last_reissued_at":"2026-05-26T01:03:12.705951Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T01:03:12.705951Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"InfiFPO: Implicit Model Fusion via Preference Optimization in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Fei Wu, Hongxia Yang, Qi Zhou, Yanggan Gu, Yiming Zhang, Yuanyi Wang, Zhaoyi Yan","submitted_at":"2025-05-20T03:32:37Z","abstract_excerpt":"Model fusion combines multiple Large Language Models (LLMs) with different strengths into a more powerful, integrated model through lightweight training methods. Existing works on model fusion focus primarily on supervised fine-tuning (SFT), leaving preference alignment (PA) --a critical phase for enhancing LLM performance--largely unexplored. The current few fusion methods on PA phase, like WRPO, simplify the process by utilizing only response outputs from source models while discarding their probability information. To address this limitation, we propose InfiFPO, a preference optimization me"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.13878","kind":"arxiv","version":3},"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/2505.13878/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":"2505.13878","created_at":"2026-05-26T01:03:12.706019+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.13878v3","created_at":"2026-05-26T01:03:12.706019+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.13878","created_at":"2026-05-26T01:03:12.706019+00:00"},{"alias_kind":"pith_short_12","alias_value":"UTCXYALO7OUI","created_at":"2026-05-26T01:03:12.706019+00:00"},{"alias_kind":"pith_short_16","alias_value":"UTCXYALO7OUIRATA","created_at":"2026-05-26T01:03:12.706019+00:00"},{"alias_kind":"pith_short_8","alias_value":"UTCXYALO","created_at":"2026-05-26T01:03:12.706019+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.16882","citing_title":"E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2509.24244","citing_title":"Model Merging Scaling Laws in Large Language Models","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14546","citing_title":"Discovering Physical Directions in Weight Space: Composing Neural PDE Experts","ref_index":42,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13030","citing_title":"FeatCal: Feature Calibration for Post-Merging Models","ref_index":60,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09608","citing_title":"Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training","ref_index":69,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU","json":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU.json","graph_json":"https://pith.science/api/pith-number/UTCXYALO7OUIRATAIH6MDJIFZU/graph.json","events_json":"https://pith.science/api/pith-number/UTCXYALO7OUIRATAIH6MDJIFZU/events.json","paper":"https://pith.science/paper/UTCXYALO"},"agent_actions":{"view_html":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU","download_json":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU.json","view_paper":"https://pith.science/paper/UTCXYALO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.13878&json=true","fetch_graph":"https://pith.science/api/pith-number/UTCXYALO7OUIRATAIH6MDJIFZU/graph.json","fetch_events":"https://pith.science/api/pith-number/UTCXYALO7OUIRATAIH6MDJIFZU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU/action/storage_attestation","attest_author":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU/action/author_attestation","sign_citation":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU/action/citation_signature","submit_replication":"https://pith.science/pith/UTCXYALO7OUIRATAIH6MDJIFZU/action/replication_record"}},"created_at":"2026-05-26T01:03:12.706019+00:00","updated_at":"2026-05-26T01:03:12.706019+00:00"}