{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RX3D7XULQUJWZAYWX56LCEJNH3","short_pith_number":"pith:RX3D7XUL","schema_version":"1.0","canonical_sha256":"8df63fde8b85136c8316bf7cb1112d3ed821e70bc2b3827439af759001b5910f","source":{"kind":"arxiv","id":"2403.07378","version":5},"attestation_state":"computed","paper":{"title":"SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Mi Zhang, Xin Wang, Yu Zheng, Zhongwei Wan","submitted_at":"2024-03-12T07:31:18Z","abstract_excerpt":"The advancements in Large Language Models (LLMs) have been hindered by their substantial sizes, which necessitates LLM compression methods for practical deployment. Singular Value Decomposition (SVD) offers a promising solution for LLM compression. However, state-of-the-art SVD-based LLM compression methods have two key limitations: truncating smaller singular values may lead to higher compression loss, and the lack of update on the compressed weights after SVD truncation. In this work, we propose SVD-LLM, a SVD-based post-training LLM compression method that addresses the limitations of exist"},"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":"2403.07378","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-03-12T07:31:18Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d99419bd6e08fe483c79ac26313cce0fba025ccf64ea13bb63bdec525a11cbcf","abstract_canon_sha256":"f9d338e2a80e11feec71a89641a783efbee79421c9a8bb0d3377b14d36c90da0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:31:42.061949Z","signature_b64":"fx9qDkISIPQyE1YtDAPH3c5zClfrMeAwJgRjl9zoibk6BgoPTK/hjlxA9cIZokbeJSEH4TbHIU8IVysdBCrVAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8df63fde8b85136c8316bf7cb1112d3ed821e70bc2b3827439af759001b5910f","last_reissued_at":"2026-07-05T10:31:42.060883Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:31:42.060883Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SVD-LLM: Truncation-aware Singular Value Decomposition for Large Language Model Compression","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Mi Zhang, Xin Wang, Yu Zheng, Zhongwei Wan","submitted_at":"2024-03-12T07:31:18Z","abstract_excerpt":"The advancements in Large Language Models (LLMs) have been hindered by their substantial sizes, which necessitates LLM compression methods for practical deployment. Singular Value Decomposition (SVD) offers a promising solution for LLM compression. However, state-of-the-art SVD-based LLM compression methods have two key limitations: truncating smaller singular values may lead to higher compression loss, and the lack of update on the compressed weights after SVD truncation. In this work, we propose SVD-LLM, a SVD-based post-training LLM compression method that addresses the limitations of exist"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.07378","kind":"arxiv","version":5},"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/2403.07378/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":"2403.07378","created_at":"2026-07-05T10:31:42.061022+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.07378v5","created_at":"2026-07-05T10:31:42.061022+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.07378","created_at":"2026-07-05T10:31:42.061022+00:00"},{"alias_kind":"pith_short_12","alias_value":"RX3D7XULQUJW","created_at":"2026-07-05T10:31:42.061022+00:00"},{"alias_kind":"pith_short_16","alias_value":"RX3D7XULQUJWZAYW","created_at":"2026-07-05T10:31:42.061022+00:00"},{"alias_kind":"pith_short_8","alias_value":"RX3D7XUL","created_at":"2026-07-05T10:31:42.061022+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":22,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19993","citing_title":"Activation- and Influence-Aware Ranks (AIR): Function-Preserving SVD Compression for LLMs","ref_index":200,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08565","citing_title":"EinSort: Sorting is All We Need for Tensorizing LLM","ref_index":84,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07098","citing_title":"SigmaScale: LLM Compression with SVD-based Low-Rank Decomposition and Learned Scaling Matrices","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03465","citing_title":"Rethinking the Role of Tensor Decompositions in Post-Training LLM Compression","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30836","citing_title":"Cross-Layer Subspace Coupling for LLM Compression: A Unifying Framework and Its Empirical Limits","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00573","citing_title":"LASER: Loss-Aware Singular-value Decomposition and Rank Allocation for Efficient Low-Precision Vision-Language Models","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.00535","citing_title":"DREAM-S: Speculative Decoding with Searchable Drafting and Target-Aware Refinement for Multimodal Generation","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2505.12942","citing_title":"A3 : an Analytical Low-Rank Approximation Framework for Attention","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18904","citing_title":"Dynamic Model Merging Made Slim","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2312.05821","citing_title":"ASVD: Activation-aware Singular Value Decomposition for Compressing Large Language Models","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17985","citing_title":"SAFE-SVD: Sensitivity-Aware Fidelity-Enforcing SVD for Physics Foundation Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17757","citing_title":"OSCAR: Offline Spectral Covariance-Aware Rotation for 2-bit KV Cache Quantization","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19842","citing_title":"Fast Tensorization of Neural Networks via Slice-wise Feature Distillation","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2404.14294","citing_title":"A Survey on Efficient Inference for Large Language Models","ref_index":229,"is_internal_anchor":false},{"citing_arxiv_id":"2604.03298","citing_title":"ENEC: A Lossless AI Model Compression Method Enabling Fast Inference on Ascend NPUs","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2604.00733","citing_title":"Spectral Compact Training: Pre-Training Large Language Models via Permanent Truncated SVD and Stiefel QR Retraction","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04013","citing_title":"RUQuant: Towards Refining Uniform Quantization for Large Language Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03109","citing_title":"Gated Subspace Inference for Transformer Acceleration","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08568","citing_title":"Different Prompts, Different Ranks: Prompt-aware Dynamic Rank Selection for SVD-based LLM Compression","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00140","citing_title":"Technical Report: Activation Residual Hessian Quantization (ARHQ) for Low-Bit LLM Quantization","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07706","citing_title":"Bayesian Fine-tuning in Projected Subspaces","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04493","citing_title":"SLaB: Sparse-Lowrank-Binary Decomposition for Efficient Large Language Models","ref_index":15,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3","json":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3.json","graph_json":"https://pith.science/api/pith-number/RX3D7XULQUJWZAYWX56LCEJNH3/graph.json","events_json":"https://pith.science/api/pith-number/RX3D7XULQUJWZAYWX56LCEJNH3/events.json","paper":"https://pith.science/paper/RX3D7XUL"},"agent_actions":{"view_html":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3","download_json":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3.json","view_paper":"https://pith.science/paper/RX3D7XUL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.07378&json=true","fetch_graph":"https://pith.science/api/pith-number/RX3D7XULQUJWZAYWX56LCEJNH3/graph.json","fetch_events":"https://pith.science/api/pith-number/RX3D7XULQUJWZAYWX56LCEJNH3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3/action/storage_attestation","attest_author":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3/action/author_attestation","sign_citation":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3/action/citation_signature","submit_replication":"https://pith.science/pith/RX3D7XULQUJWZAYWX56LCEJNH3/action/replication_record"}},"created_at":"2026-07-05T10:31:42.061022+00:00","updated_at":"2026-07-05T10:31:42.061022+00:00"}