{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:24BZEVULILIKMWPFTQEQD4RFJL","short_pith_number":"pith:24BZEVUL","schema_version":"1.0","canonical_sha256":"d70392568b42d0a659e59c0901f2254ae173372ee951aad43bb3c9ca4731b783","source":{"kind":"arxiv","id":"2412.14363","version":2},"attestation_state":"computed","paper":{"title":"ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Kaushik Roy, Sayeh Sharify, Utkarsh Saxena, Xin Wang","submitted_at":"2024-12-18T22:01:55Z","abstract_excerpt":"Post-training quantization (PTQ) of large language models (LLMs) holds the promise in reducing the prohibitive computational cost at inference time. Quantization of all weight, activation and key-value (KV) cache tensors to 4-bit without significantly degrading generalizability is challenging, due to the high quantization error caused by extreme outliers in activations. To tackle this problem, we propose ResQ, a PTQ method that pushes further the state-of-the-art. By means of principal component analysis (PCA), it identifies a low-rank subspace (in practice 1/8 of the hidden dimension) in whic"},"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":"2412.14363","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-18T22:01:55Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6902a33e413b2157f63e9a47c07d8b72c58f04db16def807929c3cc3e09e0565","abstract_canon_sha256":"eb76137d0dd1c7570163e54669b2740d3cfb291cc397b83c23b4c16c13687789"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:09:13.970278Z","signature_b64":"es/1ut7ypti9JcOqVT+d0FLfPS3w+EDAq+MBONNj/G73dtmNamcDIZoKIKmmUUjLbtCR7MkSaMm54LI+t1yKBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d70392568b42d0a659e59c0901f2254ae173372ee951aad43bb3c9ca4731b783","last_reissued_at":"2026-07-05T10:09:13.969791Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:09:13.969791Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Kaushik Roy, Sayeh Sharify, Utkarsh Saxena, Xin Wang","submitted_at":"2024-12-18T22:01:55Z","abstract_excerpt":"Post-training quantization (PTQ) of large language models (LLMs) holds the promise in reducing the prohibitive computational cost at inference time. Quantization of all weight, activation and key-value (KV) cache tensors to 4-bit without significantly degrading generalizability is challenging, due to the high quantization error caused by extreme outliers in activations. To tackle this problem, we propose ResQ, a PTQ method that pushes further the state-of-the-art. By means of principal component analysis (PCA), it identifies a low-rank subspace (in practice 1/8 of the hidden dimension) in whic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.14363","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/2412.14363/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":"2412.14363","created_at":"2026-07-05T10:09:13.969851+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.14363v2","created_at":"2026-07-05T10:09:13.969851+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.14363","created_at":"2026-07-05T10:09:13.969851+00:00"},{"alias_kind":"pith_short_12","alias_value":"24BZEVULILIK","created_at":"2026-07-05T10:09:13.969851+00:00"},{"alias_kind":"pith_short_16","alias_value":"24BZEVULILIKMWPF","created_at":"2026-07-05T10:09:13.969851+00:00"},{"alias_kind":"pith_short_8","alias_value":"24BZEVUL","created_at":"2026-07-05T10:09:13.969851+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26587","citing_title":"SharQ: Bridging Activation Sparsity and FP4 Quantization for LLM Inference","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12876","citing_title":"Multi-Bitwidth Quantization for LLMs Using Additive Codebooks","ref_index":92,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04115","citing_title":"dMX: Differentiable Mixed-Precision Assignment for Low-Precision Floating-Point Formats","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20682","citing_title":"Variance Is Not Importance: Structural Analysis of Transformer Compressibility Across Model Scales","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL","json":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL.json","graph_json":"https://pith.science/api/pith-number/24BZEVULILIKMWPFTQEQD4RFJL/graph.json","events_json":"https://pith.science/api/pith-number/24BZEVULILIKMWPFTQEQD4RFJL/events.json","paper":"https://pith.science/paper/24BZEVUL"},"agent_actions":{"view_html":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL","download_json":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL.json","view_paper":"https://pith.science/paper/24BZEVUL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.14363&json=true","fetch_graph":"https://pith.science/api/pith-number/24BZEVULILIKMWPFTQEQD4RFJL/graph.json","fetch_events":"https://pith.science/api/pith-number/24BZEVULILIKMWPFTQEQD4RFJL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL/action/storage_attestation","attest_author":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL/action/author_attestation","sign_citation":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL/action/citation_signature","submit_replication":"https://pith.science/pith/24BZEVULILIKMWPFTQEQD4RFJL/action/replication_record"}},"created_at":"2026-07-05T10:09:13.969851+00:00","updated_at":"2026-07-05T10:09:13.969851+00:00"}