{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QYXF3TD2L4UKPZPTLB73ZC26PG","short_pith_number":"pith:QYXF3TD2","schema_version":"1.0","canonical_sha256":"862e5dcc7a5f28a7e5f3587fbc8b5e79a91b5f98af3fcc71e50b429830f3ac12","source":{"kind":"arxiv","id":"2412.07380","version":3},"attestation_state":"computed","paper":{"title":"SpecFuse: Ensembling Large Language Models via Next-Segment Prediction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bo Lv, Chen Tang, Nayu Liu, Ping Luo, Xin Liu, Yue Yu","submitted_at":"2024-12-10T10:27:41Z","abstract_excerpt":"Ensembles of generative large language models (LLMs) are a promising way to compensate for individual model limitations, integrating the strengths of different LLMs. Existing LLM ensemble methods, however, face limitations such as first-token delay and challenges in long-range semantic collaboration between models, Moreover, they typically assume equal voting weights for all models during ensemble, ignoring task-specific performance differences among models. In this work, we propose SpecEM, a training-free, plug-and-play LLM ensemble framework that dynamically adjusts each model's model contri"},"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":"2412.07380","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-12-10T10:27:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"97139648342f47a6131fe0615068dad3310f38ef5df40511932199aca1e1783f","abstract_canon_sha256":"ebfcb62c9f7d1a5de5d67ea342835d8b73766c42b35ae7df36c144035ffd34f6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-29T01:14:01.168006Z","signature_b64":"I327x+J9gCVlHAcIwBwxme6g1xJmV89YOSGXX9N237VbywrX+4/UmdyKh/493pokknITHcUJLvCf3jsAQyDUBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"862e5dcc7a5f28a7e5f3587fbc8b5e79a91b5f98af3fcc71e50b429830f3ac12","last_reissued_at":"2026-06-29T01:14:01.167475Z","signature_status":"signed_v1","first_computed_at":"2026-06-29T01:14:01.167475Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SpecFuse: Ensembling Large Language Models via Next-Segment Prediction","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bo Lv, Chen Tang, Nayu Liu, Ping Luo, Xin Liu, Yue Yu","submitted_at":"2024-12-10T10:27:41Z","abstract_excerpt":"Ensembles of generative large language models (LLMs) are a promising way to compensate for individual model limitations, integrating the strengths of different LLMs. Existing LLM ensemble methods, however, face limitations such as first-token delay and challenges in long-range semantic collaboration between models, Moreover, they typically assume equal voting weights for all models during ensemble, ignoring task-specific performance differences among models. In this work, we propose SpecEM, a training-free, plug-and-play LLM ensemble framework that dynamically adjusts each model's model contri"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.07380","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/2412.07380/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":"2412.07380","created_at":"2026-06-29T01:14:01.167538+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.07380v3","created_at":"2026-06-29T01:14:01.167538+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.07380","created_at":"2026-06-29T01:14:01.167538+00:00"},{"alias_kind":"pith_short_12","alias_value":"QYXF3TD2L4UK","created_at":"2026-06-29T01:14:01.167538+00:00"},{"alias_kind":"pith_short_16","alias_value":"QYXF3TD2L4UKPZPT","created_at":"2026-06-29T01:14:01.167538+00:00"},{"alias_kind":"pith_short_8","alias_value":"QYXF3TD2","created_at":"2026-06-29T01:14:01.167538+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2502.18036","citing_title":"Harnessing Multiple Large Language Models: A Survey on LLM Ensemble","ref_index":30,"is_internal_anchor":true},{"citing_arxiv_id":"2506.14123","citing_title":"Sampling from Your Language Model One Byte at a Time","ref_index":40,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG","json":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG.json","graph_json":"https://pith.science/api/pith-number/QYXF3TD2L4UKPZPTLB73ZC26PG/graph.json","events_json":"https://pith.science/api/pith-number/QYXF3TD2L4UKPZPTLB73ZC26PG/events.json","paper":"https://pith.science/paper/QYXF3TD2"},"agent_actions":{"view_html":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG","download_json":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG.json","view_paper":"https://pith.science/paper/QYXF3TD2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.07380&json=true","fetch_graph":"https://pith.science/api/pith-number/QYXF3TD2L4UKPZPTLB73ZC26PG/graph.json","fetch_events":"https://pith.science/api/pith-number/QYXF3TD2L4UKPZPTLB73ZC26PG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG/action/storage_attestation","attest_author":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG/action/author_attestation","sign_citation":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG/action/citation_signature","submit_replication":"https://pith.science/pith/QYXF3TD2L4UKPZPTLB73ZC26PG/action/replication_record"}},"created_at":"2026-06-29T01:14:01.167538+00:00","updated_at":"2026-06-29T01:14:01.167538+00:00"}