{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:UNRGCCT6OZRN5ZMR4WWZNT7DMC","short_pith_number":"pith:UNRGCCT6","schema_version":"1.0","canonical_sha256":"a362610a7e7662dee591e5ad96cfe360bda632dc2751d920286ab0b3e921b026","source":{"kind":"arxiv","id":"2306.02561","version":3},"attestation_state":"computed","paper":{"title":"LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bill Yuchen Lin, Dongfu Jiang, Xiang Ren","submitted_at":"2023-06-05T03:32:26Z","abstract_excerpt":"We present LLM-Blender, an ensembling framework designed to attain consistently superior performance by leveraging the diverse strengths of multiple open-source large language models (LLMs). Our framework consists of two modules: PairRanker and GenFuser, addressing the observation that optimal LLMs for different examples can significantly vary. PairRanker employs a specialized pairwise comparison method to distinguish subtle differences between candidate outputs. It jointly encodes the input text and a pair of candidates, using cross-attention encoders to determine the superior one. Our result"},"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":"2306.02561","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-06-05T03:32:26Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"a5e52fd9da493139d93477537315665cd4947522f80055c0c319fde3850a6bc1","abstract_canon_sha256":"cf9f38150590cf97e3088843bfb3e75124a7b11af741013ab1e6cbf4f9abf7e5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:26:50.117453Z","signature_b64":"ELZkBS4qomlE3ra4Iawyj2Jmh/65Uz1Euzf2ZTLwx95zQFYHm4jFP1MPbXlYn/FRKEwMAf6c2zTcG2EIdAt8AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a362610a7e7662dee591e5ad96cfe360bda632dc2751d920286ab0b3e921b026","last_reissued_at":"2026-07-05T06:26:50.116996Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:26:50.116996Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bill Yuchen Lin, Dongfu Jiang, Xiang Ren","submitted_at":"2023-06-05T03:32:26Z","abstract_excerpt":"We present LLM-Blender, an ensembling framework designed to attain consistently superior performance by leveraging the diverse strengths of multiple open-source large language models (LLMs). Our framework consists of two modules: PairRanker and GenFuser, addressing the observation that optimal LLMs for different examples can significantly vary. PairRanker employs a specialized pairwise comparison method to distinguish subtle differences between candidate outputs. It jointly encodes the input text and a pair of candidates, using cross-attention encoders to determine the superior one. Our result"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.02561","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/2306.02561/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":"2306.02561","created_at":"2026-07-05T06:26:50.117051+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.02561v3","created_at":"2026-07-05T06:26:50.117051+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.02561","created_at":"2026-07-05T06:26:50.117051+00:00"},{"alias_kind":"pith_short_12","alias_value":"UNRGCCT6OZRN","created_at":"2026-07-05T06:26:50.117051+00:00"},{"alias_kind":"pith_short_16","alias_value":"UNRGCCT6OZRN5ZMR","created_at":"2026-07-05T06:26:50.117051+00:00"},{"alias_kind":"pith_short_8","alias_value":"UNRGCCT6","created_at":"2026-07-05T06:26:50.117051+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":19,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25476","citing_title":"A Red Teaming Framework for Large Language Models: A Case Study on Faithfulness Evaluation","ref_index":80,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27288","citing_title":"When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18774","citing_title":"RouteJudge: An Open Platform for Reproducible and Preference-Aware LLM Routing","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00053","citing_title":"SWE-Router: Routing in Multi-turn Agentic Software Engineering Tasks","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07392","citing_title":"Online Pandora's Box for Contextual LLM Cascading","ref_index":70,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27810","citing_title":"LRanker: LLM Ranker for Massive Candidates","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2505.12601","citing_title":"Rethinking Predictive Modeling for LLM Routing: When Simple kNN Beats Complex Learned Routers","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2507.14200","citing_title":"A Scalable Multi-LLM Collaboration System with Retrieval-based Selection and Exploration-Exploitation-Driven Enhancement","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15207","citing_title":"TeamTR: Trust-Region Fine-Tuning for Multi-Agent LLM Coordination","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2410.18451","citing_title":"Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs","ref_index":11,"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":27,"is_internal_anchor":false},{"citing_arxiv_id":"2602.02350","citing_title":"Context Learning for Multi-Agent Discussion","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2407.21787","citing_title":"Large Language Monkeys: Scaling Inference Compute with Repeated Sampling","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2406.18665","citing_title":"RouteLLM: Learning to Route LLMs with Preference Data","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10023","citing_title":"FREE-Switch: Frequency-based Dynamic LoRA Switch for Style Transfer","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11141","citing_title":"Reducing Hallucination in Enterprise AI Workflows via Hybrid Utility Minimum Bayes Risk (HUMBR)","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06987","citing_title":"Response Time Enhances Alignment with Heterogeneous Preferences","ref_index":91,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15728","citing_title":"Privacy-Preserving LLMs Routing","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17950","citing_title":"CADMAS-CTX: Contextual Capability Calibration for Multi-Agent Delegation","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC","json":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC.json","graph_json":"https://pith.science/api/pith-number/UNRGCCT6OZRN5ZMR4WWZNT7DMC/graph.json","events_json":"https://pith.science/api/pith-number/UNRGCCT6OZRN5ZMR4WWZNT7DMC/events.json","paper":"https://pith.science/paper/UNRGCCT6"},"agent_actions":{"view_html":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC","download_json":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC.json","view_paper":"https://pith.science/paper/UNRGCCT6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.02561&json=true","fetch_graph":"https://pith.science/api/pith-number/UNRGCCT6OZRN5ZMR4WWZNT7DMC/graph.json","fetch_events":"https://pith.science/api/pith-number/UNRGCCT6OZRN5ZMR4WWZNT7DMC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC/action/storage_attestation","attest_author":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC/action/author_attestation","sign_citation":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC/action/citation_signature","submit_replication":"https://pith.science/pith/UNRGCCT6OZRN5ZMR4WWZNT7DMC/action/replication_record"}},"created_at":"2026-07-05T06:26:50.117051+00:00","updated_at":"2026-07-05T06:26:50.117051+00:00"}