{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TZVVFFCHDMBRW7Z55SAXQZIITR","short_pith_number":"pith:TZVVFFCH","schema_version":"1.0","canonical_sha256":"9e6b5294471b031b7f3dec817865089c61ec5444f96ec8bd9b202f17f81f1404","source":{"kind":"arxiv","id":"2407.19807","version":2},"attestation_state":"computed","paper":{"title":"Cool-Fusion: Fuse Large Language Models without Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cong Liu, Liang Lin, Weigang Wu, Xiaojun Quan, Xu Chen, Yan Pan","submitted_at":"2024-07-29T09:02:19Z","abstract_excerpt":"We focus on the problem of fusing two or more heterogeneous large language models (LLMs) to leverage their complementary strengths. One of the challenges of model fusion is high computational load, specifically in fine-tuning or aligning vocabularies. To address this, we propose Cool-Fusion, a simple yet effective approach that fuses the knowledge of source LLMs, which does not require training. Unlike ensemble methods, Cool-Fusion is applicable to any set of source LLMs that have different vocabularies. To overcome the vocabulary discrepancies among LLMs, we ensemble LLMs on text level, allow"},"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":"2407.19807","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-29T09:02:19Z","cross_cats_sorted":[],"title_canon_sha256":"198e1e7789320987b385cf31a5f4fed6e033286ed7b6d5c8a719acc9a8db4890","abstract_canon_sha256":"7e9e1dfc020f655c8275b841bb0f12a6f1a6b4bc3e8111f32b09520f6ad527ce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:18:13.564862Z","signature_b64":"ivN/ZNce9JG7awDukuy5LdDazym3NuEn0wh1JdNif/7y9P/LFrQP3TzbmTDSbodDwLmWzCJlRq9LrfLvb509Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9e6b5294471b031b7f3dec817865089c61ec5444f96ec8bd9b202f17f81f1404","last_reissued_at":"2026-07-05T11:18:13.564389Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:18:13.564389Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cool-Fusion: Fuse Large Language Models without Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Cong Liu, Liang Lin, Weigang Wu, Xiaojun Quan, Xu Chen, Yan Pan","submitted_at":"2024-07-29T09:02:19Z","abstract_excerpt":"We focus on the problem of fusing two or more heterogeneous large language models (LLMs) to leverage their complementary strengths. One of the challenges of model fusion is high computational load, specifically in fine-tuning or aligning vocabularies. To address this, we propose Cool-Fusion, a simple yet effective approach that fuses the knowledge of source LLMs, which does not require training. Unlike ensemble methods, Cool-Fusion is applicable to any set of source LLMs that have different vocabularies. To overcome the vocabulary discrepancies among LLMs, we ensemble LLMs on text level, allow"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.19807","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/2407.19807/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":"2407.19807","created_at":"2026-07-05T11:18:13.564447+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.19807v2","created_at":"2026-07-05T11:18:13.564447+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.19807","created_at":"2026-07-05T11:18:13.564447+00:00"},{"alias_kind":"pith_short_12","alias_value":"TZVVFFCHDMBR","created_at":"2026-07-05T11:18:13.564447+00:00"},{"alias_kind":"pith_short_16","alias_value":"TZVVFFCHDMBRW7Z5","created_at":"2026-07-05T11:18:13.564447+00:00"},{"alias_kind":"pith_short_8","alias_value":"TZVVFFCH","created_at":"2026-07-05T11:18:13.564447+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2502.18036","citing_title":"Harnessing Multiple Large Language Models: A Survey on LLM Ensemble","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2506.14123","citing_title":"Sampling from Your Language Model One Byte at a Time","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2512.23213","citing_title":"Scoring, Reasoning, and Selecting the Best! Ensembling Large Language Models via a Peer-Review Process","ref_index":35,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR","json":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR.json","graph_json":"https://pith.science/api/pith-number/TZVVFFCHDMBRW7Z55SAXQZIITR/graph.json","events_json":"https://pith.science/api/pith-number/TZVVFFCHDMBRW7Z55SAXQZIITR/events.json","paper":"https://pith.science/paper/TZVVFFCH"},"agent_actions":{"view_html":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR","download_json":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR.json","view_paper":"https://pith.science/paper/TZVVFFCH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.19807&json=true","fetch_graph":"https://pith.science/api/pith-number/TZVVFFCHDMBRW7Z55SAXQZIITR/graph.json","fetch_events":"https://pith.science/api/pith-number/TZVVFFCHDMBRW7Z55SAXQZIITR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR/action/storage_attestation","attest_author":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR/action/author_attestation","sign_citation":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR/action/citation_signature","submit_replication":"https://pith.science/pith/TZVVFFCHDMBRW7Z55SAXQZIITR/action/replication_record"}},"created_at":"2026-07-05T11:18:13.564447+00:00","updated_at":"2026-07-05T11:18:13.564447+00:00"}