{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:UQAWMO4KTVSHOE3S6KTO3XGONA","short_pith_number":"pith:UQAWMO4K","schema_version":"1.0","canonical_sha256":"a401663b8a9d64771372f2a6eddcce681edc2b855dfb3d56e852f34a1350bda8","source":{"kind":"arxiv","id":"2510.26690","version":3},"attestation_state":"computed","paper":{"title":"LoRAQuant: Mixed-Precision Quantization of LoRA to Ultra-Low Bits","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amir Reza Mirzaei, Lili Mou, Yanshuai Cao, Yuqiao Wen","submitted_at":"2025-10-30T16:59:22Z","abstract_excerpt":"Low-Rank Adaptation (LoRA) has become a popular technique for parameter-efficient fine-tuning of large language models (LLMs). In many real-world scenarios, multiple adapters are loaded simultaneously to enable LLM customization for personalized user experiences or to support a diverse range of tasks. Although each adapter is lightweight in isolation, their aggregate cost becomes substantial at scale. To address this, we propose LoRAQuant, a mixed-precision post-training quantization method tailored to LoRA. Specifically, LoRAQuant reparameterizes each adapter by singular value decomposition ("},"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":"2510.26690","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-10-30T16:59:22Z","cross_cats_sorted":[],"title_canon_sha256":"4b333c398ed5407a26c6f284ebc359cc53078dea2617e1928c8682262c94e4da","abstract_canon_sha256":"e158a88dbdc77a9e4a2742b8a5b89f1dc6e4e10feba6a9babeaefea71fb9cc54"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-13T00:21:14.234078Z","signature_b64":"t0cHYJliwxx44tHvRBW9u9AG/w+TD/xtsn0FcGisi1lgzI9II2S2vs59pnOI/FHEPKUbQblEgp+RW7tqlwXAAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a401663b8a9d64771372f2a6eddcce681edc2b855dfb3d56e852f34a1350bda8","last_reissued_at":"2026-08-13T00:21:14.232412Z","signature_status":"signed_v1","first_computed_at":"2026-08-13T00:21:14.232412Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoRAQuant: Mixed-Precision Quantization of LoRA to Ultra-Low Bits","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Amir Reza Mirzaei, Lili Mou, Yanshuai Cao, Yuqiao Wen","submitted_at":"2025-10-30T16:59:22Z","abstract_excerpt":"Low-Rank Adaptation (LoRA) has become a popular technique for parameter-efficient fine-tuning of large language models (LLMs). In many real-world scenarios, multiple adapters are loaded simultaneously to enable LLM customization for personalized user experiences or to support a diverse range of tasks. Although each adapter is lightweight in isolation, their aggregate cost becomes substantial at scale. To address this, we propose LoRAQuant, a mixed-precision post-training quantization method tailored to LoRA. Specifically, LoRAQuant reparameterizes each adapter by singular value decomposition ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.26690","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/2510.26690/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":"2510.26690","created_at":"2026-08-13T00:21:14.232742+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.26690v3","created_at":"2026-08-13T00:21:14.232742+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.26690","created_at":"2026-08-13T00:21:14.232742+00:00"},{"alias_kind":"pith_short_12","alias_value":"UQAWMO4KTVSH","created_at":"2026-08-13T00:21:14.232742+00:00"},{"alias_kind":"pith_short_16","alias_value":"UQAWMO4KTVSHOE3S","created_at":"2026-08-13T00:21:14.232742+00:00"},{"alias_kind":"pith_short_8","alias_value":"UQAWMO4K","created_at":"2026-08-13T00:21:14.232742+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.24058","citing_title":"Signs Beat Floats: Low-Rank Double-Binary Adaptation for On-Device Fine-Tuning","ref_index":30,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA","json":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA.json","graph_json":"https://pith.science/api/pith-number/UQAWMO4KTVSHOE3S6KTO3XGONA/graph.json","events_json":"https://pith.science/api/pith-number/UQAWMO4KTVSHOE3S6KTO3XGONA/events.json","paper":"https://pith.science/paper/UQAWMO4K"},"agent_actions":{"view_html":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA","download_json":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA.json","view_paper":"https://pith.science/paper/UQAWMO4K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.26690&json=true","fetch_graph":"https://pith.science/api/pith-number/UQAWMO4KTVSHOE3S6KTO3XGONA/graph.json","fetch_events":"https://pith.science/api/pith-number/UQAWMO4KTVSHOE3S6KTO3XGONA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA/action/storage_attestation","attest_author":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA/action/author_attestation","sign_citation":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA/action/citation_signature","submit_replication":"https://pith.science/pith/UQAWMO4KTVSHOE3S6KTO3XGONA/action/replication_record"}},"created_at":"2026-08-13T00:21:14.232742+00:00","updated_at":"2026-08-13T00:21:14.232742+00:00"}