{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:ND2WWIX3P24QGFT7KGNG5SP7UX","short_pith_number":"pith:ND2WWIX3","schema_version":"1.0","canonical_sha256":"68f56b22fb7eb903167f519a6ec9ffa5f6a9920a9965dfe00fd31c3c6dea2a6f","source":{"kind":"arxiv","id":"2309.05444","version":1},"attestation_state":"computed","paper":{"title":"Pushing Mixture of Experts to the Limit: Extremely Parameter Efficient MoE for Instruction Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Acyr Locatelli, Ahmet \\\"Ust\\\"un, Arash Ahmadian, Beyza Ermi\\c{s}, Sara Hooker, Ted Zadouri","submitted_at":"2023-09-11T13:31:00Z","abstract_excerpt":"The Mixture of Experts (MoE) is a widely known neural architecture where an ensemble of specialized sub-models optimizes overall performance with a constant computational cost. However, conventional MoEs pose challenges at scale due to the need to store all experts in memory. In this paper, we push MoE to the limit. We propose extremely parameter-efficient MoE by uniquely combining MoE architecture with lightweight experts.Our MoE architecture outperforms standard parameter-efficient fine-tuning (PEFT) methods and is on par with full fine-tuning by only updating the lightweight experts -- less"},"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":"2309.05444","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-11T13:31:00Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"e68171099625f2d28d883458472182858c17233a63a56044aa1bb1d49734300f","abstract_canon_sha256":"958496c0763840523c1e33319c2564e4bade7e0a404e8c5ff5e5a413317c7b9f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:49:35.358145Z","signature_b64":"l+ZY0rUhGpl75kQ32l13HhSjSqsGL1g2vFTk+4VLnFSU1M+N+aNYw4bIL0Ll65Eq9XCdt3zjzxB1wP1L/taZCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"68f56b22fb7eb903167f519a6ec9ffa5f6a9920a9965dfe00fd31c3c6dea2a6f","last_reissued_at":"2026-07-05T06:49:35.357524Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:49:35.357524Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pushing Mixture of Experts to the Limit: Extremely Parameter Efficient MoE for Instruction Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Acyr Locatelli, Ahmet \\\"Ust\\\"un, Arash Ahmadian, Beyza Ermi\\c{s}, Sara Hooker, Ted Zadouri","submitted_at":"2023-09-11T13:31:00Z","abstract_excerpt":"The Mixture of Experts (MoE) is a widely known neural architecture where an ensemble of specialized sub-models optimizes overall performance with a constant computational cost. However, conventional MoEs pose challenges at scale due to the need to store all experts in memory. In this paper, we push MoE to the limit. We propose extremely parameter-efficient MoE by uniquely combining MoE architecture with lightweight experts.Our MoE architecture outperforms standard parameter-efficient fine-tuning (PEFT) methods and is on par with full fine-tuning by only updating the lightweight experts -- less"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.05444","kind":"arxiv","version":1},"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/2309.05444/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":"2309.05444","created_at":"2026-07-05T06:49:35.357590+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.05444v1","created_at":"2026-07-05T06:49:35.357590+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.05444","created_at":"2026-07-05T06:49:35.357590+00:00"},{"alias_kind":"pith_short_12","alias_value":"ND2WWIX3P24Q","created_at":"2026-07-05T06:49:35.357590+00:00"},{"alias_kind":"pith_short_16","alias_value":"ND2WWIX3P24QGFT7","created_at":"2026-07-05T06:49:35.357590+00:00"},{"alias_kind":"pith_short_8","alias_value":"ND2WWIX3","created_at":"2026-07-05T06:49:35.357590+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21645","citing_title":"Behavioral and Representational Evidence of Binomial Ordering Preferences in Large Language Models","ref_index":123,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19079","citing_title":"ARIADNE: Agnostic Routing for Inference-time Adapter DyNamic sElection","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26376","citing_title":"BioFact-MoE: Biologically Factorized Mixture of Experts for Vision-Language Prognostic Modeling in Hepatocellular Carcinoma","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20247","citing_title":"CP-MoE: Consistency-Preserving Mixture-of-Experts for Continual Learning","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2603.18297","citing_title":"Path-Constrained Mixture-of-Experts","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2403.14608","citing_title":"Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26340","citing_title":"Adaptive and Fine-grained Module-wise Expert Pruning for Efficient LoRA-MoE Fine-Tuning","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19835","citing_title":"Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20721","citing_title":"ALAS: Adaptive Long-Horizon Action Synthesis via Async-pathway Stream Disentanglement","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04079","citing_title":"Efficient Handwriting-Based Alzheimer,s Disease Diagnosis Using a Low-Rank Mixture of Experts Deep Learning Framework","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19835","citing_title":"Expert Upcycling: Shifting the Compute-Efficient Frontier of Mixture-of-Experts","ref_index":57,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX","json":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX.json","graph_json":"https://pith.science/api/pith-number/ND2WWIX3P24QGFT7KGNG5SP7UX/graph.json","events_json":"https://pith.science/api/pith-number/ND2WWIX3P24QGFT7KGNG5SP7UX/events.json","paper":"https://pith.science/paper/ND2WWIX3"},"agent_actions":{"view_html":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX","download_json":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX.json","view_paper":"https://pith.science/paper/ND2WWIX3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.05444&json=true","fetch_graph":"https://pith.science/api/pith-number/ND2WWIX3P24QGFT7KGNG5SP7UX/graph.json","fetch_events":"https://pith.science/api/pith-number/ND2WWIX3P24QGFT7KGNG5SP7UX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX/action/storage_attestation","attest_author":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX/action/author_attestation","sign_citation":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX/action/citation_signature","submit_replication":"https://pith.science/pith/ND2WWIX3P24QGFT7KGNG5SP7UX/action/replication_record"}},"created_at":"2026-07-05T06:49:35.357590+00:00","updated_at":"2026-07-05T06:49:35.357590+00:00"}