{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:6XMQUCC4SSZSGTYVXVXOQ5K5X4","short_pith_number":"pith:6XMQUCC4","schema_version":"1.0","canonical_sha256":"f5d90a085c94b3234f15bd6ee8755dbf382724bf527ee8519f36967bcb8676d7","source":{"kind":"arxiv","id":"2301.00774","version":3},"attestation_state":"computed","paper":{"title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dan Alistarh, Elias Frantar","submitted_at":"2023-01-02T17:48:56Z","abstract_excerpt":"We show for the first time that large-scale generative pretrained transformer (GPT) family models can be pruned to at least 50% sparsity in one-shot, without any retraining, at minimal loss of accuracy. This is achieved via a new pruning method called SparseGPT, specifically designed to work efficiently and accurately on massive GPT-family models. We can execute SparseGPT on the largest available open-source models, OPT-175B and BLOOM-176B, in under 4.5 hours, and can reach 60% unstructured sparsity with negligible increase in perplexity: remarkably, more than 100 billion weights from these mo"},"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":"2301.00774","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-01-02T17:48:56Z","cross_cats_sorted":[],"title_canon_sha256":"588bebc2f774852adb59344bd0213a60a217fbefde8c1dc6fa3efede7ff45a2b","abstract_canon_sha256":"944b3f0912312db8eb073610596bc626b8475a1e75b8f74a692b8ad96849cf8a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:53:35.560304Z","signature_b64":"BSo7PJmav/ycrMtGnHe/CdPEGb8h1krFrHFVeQx4j0xtaKay6a3+e+4D1qREp2yFg296WgLMK4HnBytO6LRECA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f5d90a085c94b3234f15bd6ee8755dbf382724bf527ee8519f36967bcb8676d7","last_reissued_at":"2026-07-05T05:53:35.559809Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:53:35.559809Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SparseGPT: Massive Language Models Can Be Accurately Pruned in One-Shot","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dan Alistarh, Elias Frantar","submitted_at":"2023-01-02T17:48:56Z","abstract_excerpt":"We show for the first time that large-scale generative pretrained transformer (GPT) family models can be pruned to at least 50% sparsity in one-shot, without any retraining, at minimal loss of accuracy. This is achieved via a new pruning method called SparseGPT, specifically designed to work efficiently and accurately on massive GPT-family models. We can execute SparseGPT on the largest available open-source models, OPT-175B and BLOOM-176B, in under 4.5 hours, and can reach 60% unstructured sparsity with negligible increase in perplexity: remarkably, more than 100 billion weights from these mo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2301.00774","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/2301.00774/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":"2301.00774","created_at":"2026-07-05T05:53:35.559868+00:00"},{"alias_kind":"arxiv_version","alias_value":"2301.00774v3","created_at":"2026-07-05T05:53:35.559868+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2301.00774","created_at":"2026-07-05T05:53:35.559868+00:00"},{"alias_kind":"pith_short_12","alias_value":"6XMQUCC4SSZS","created_at":"2026-07-05T05:53:35.559868+00:00"},{"alias_kind":"pith_short_16","alias_value":"6XMQUCC4SSZSGTYV","created_at":"2026-07-05T05:53:35.559868+00:00"},{"alias_kind":"pith_short_8","alias_value":"6XMQUCC4","created_at":"2026-07-05T05:53:35.559868+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":25,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07557","citing_title":"PALS: Percentile-Aware Layerwise Sparsity for LLM Pruning","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2606.24937","citing_title":"The Hitchhiker's Guide to Agentic AI: From Foundations to Systems","ref_index":151,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06254","citing_title":"SecRL-Prune: Structured Reinforcement Learning-Based Pruning of CodeLLMs for Preserving Adversarial Code Mutation","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04032","citing_title":"Do Transformers Need Three Projections? Systematic Study of QKV Variants","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2605.31256","citing_title":"Before Parc Ferm\\'e: RL-Time Pruning for Efficient Embodied LLMs in Autonomous Driving","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28438","citing_title":"When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15491","citing_title":"Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19350","citing_title":"Pruning via Causal Attribution Preserves Reasoning Performance in Large Language Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24754","citing_title":"Motion-Compensated Weight Compression","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.26538","citing_title":"CascadeFormer: Depth-Tapered Transformers Motivated by Gradient Fan-in Asymmetry","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2412.00069","citing_title":"Condense, Don't Just Prune: Enhancing Efficiency and Performance in MoE Layer Pruning","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2503.08223","citing_title":"Will LLMs Scaling Hit the Wall? Breaking Barriers via Distributed Resources on Massive Edge Devices","ref_index":167,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":74,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15491","citing_title":"Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2506.12876","citing_title":"MaskPro: Linear-Space Probabilistic Learning for Strict (N:M)-Sparsity on LLMs","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2509.12464","citing_title":"Reasoning Models Can be Accurately Pruned Via Chain-of-Thought Reconstruction","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2510.09696","citing_title":"Vanishing Contributions: A Unified Framework for Smooth and Iterative Model Compression","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2306.14048","citing_title":"H$_2$O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2602.01997","citing_title":"On the Limits of Layer Pruning for Generative Reasoning in Large Language Models","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2602.05243","citing_title":"CORP: Closed-Form One-shot Representation-Preserving Structured Pruning for Transformers","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14156","citing_title":"Compressed-Sensing-Guided, Inference-Aware Structured Reduction for Large Language Models","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":74,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10933","citing_title":"DECO: Sparse Mixture-of-Experts with Dense-Comparable Performance on End-Side Devices","ref_index":74,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06402","citing_title":"SparseForge: Efficient Semi-Structured LLM Sparsification via Annealing of Hessian-Guided Soft-Mask","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06440","citing_title":"Visual prompting reimagined: The power of the Activation Prompts","ref_index":4,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4","json":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4.json","graph_json":"https://pith.science/api/pith-number/6XMQUCC4SSZSGTYVXVXOQ5K5X4/graph.json","events_json":"https://pith.science/api/pith-number/6XMQUCC4SSZSGTYVXVXOQ5K5X4/events.json","paper":"https://pith.science/paper/6XMQUCC4"},"agent_actions":{"view_html":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4","download_json":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4.json","view_paper":"https://pith.science/paper/6XMQUCC4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2301.00774&json=true","fetch_graph":"https://pith.science/api/pith-number/6XMQUCC4SSZSGTYVXVXOQ5K5X4/graph.json","fetch_events":"https://pith.science/api/pith-number/6XMQUCC4SSZSGTYVXVXOQ5K5X4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4/action/storage_attestation","attest_author":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4/action/author_attestation","sign_citation":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4/action/citation_signature","submit_replication":"https://pith.science/pith/6XMQUCC4SSZSGTYVXVXOQ5K5X4/action/replication_record"}},"created_at":"2026-07-05T05:53:35.559868+00:00","updated_at":"2026-07-05T05:53:35.559868+00:00"}