{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:Q6CQSEQJDWBZTW7AXZUZIV7E5J","short_pith_number":"pith:Q6CQSEQJ","schema_version":"1.0","canonical_sha256":"87850912091d8399dbe0be699457e4ea4d3852de962b5ff9fcd648b1f3c00b2c","source":{"kind":"arxiv","id":"2411.07191","version":2},"attestation_state":"computed","paper":{"title":"The Super Weight in Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alvin Wan, Colorado J Reed, De Wang, Mengxia Yu, Qi Shan","submitted_at":"2024-11-11T18:05:48Z","abstract_excerpt":"Recent works have shown a surprising result: a small fraction of Large Language Model (LLM) parameter outliers are disproportionately important to the quality of the model. LLMs contain billions of parameters, so these small fractions, such as 0.01%, translate to hundreds of thousands of parameters. In this work, we present an even more surprising finding: Pruning as few as a single parameter can destroy an LLM's ability to generate text -- increasing perplexity by 3 orders of magnitude and reducing zero-shot accuracy to guessing. We propose a data-free method for identifying such parameters, "},"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":"2411.07191","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-11-11T18:05:48Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f9635e0e0e3ed686dc0303257cc2a1c16c5aad3dcea9a1a46812c1384f8ef7ce","abstract_canon_sha256":"7dbd492298179fcf7fed5d4e8d5e3a094621c3f6729b7bc0ee9c7f33832ea463"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:33:02.676951Z","signature_b64":"6a7Mbvi/+bvfwWmUAC5ulrAcEp/6qjZoYXARnUwNMOdQ6qNFxPe68ZGbwPkaL1Af8gRWOoesC8PRcjSwnWQBAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"87850912091d8399dbe0be699457e4ea4d3852de962b5ff9fcd648b1f3c00b2c","last_reissued_at":"2026-07-05T11:33:02.676447Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:33:02.676447Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Super Weight in Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Alvin Wan, Colorado J Reed, De Wang, Mengxia Yu, Qi Shan","submitted_at":"2024-11-11T18:05:48Z","abstract_excerpt":"Recent works have shown a surprising result: a small fraction of Large Language Model (LLM) parameter outliers are disproportionately important to the quality of the model. LLMs contain billions of parameters, so these small fractions, such as 0.01%, translate to hundreds of thousands of parameters. In this work, we present an even more surprising finding: Pruning as few as a single parameter can destroy an LLM's ability to generate text -- increasing perplexity by 3 orders of magnitude and reducing zero-shot accuracy to guessing. We propose a data-free method for identifying such parameters, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.07191","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/2411.07191/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":"2411.07191","created_at":"2026-07-05T11:33:02.676507+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.07191v2","created_at":"2026-07-05T11:33:02.676507+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.07191","created_at":"2026-07-05T11:33:02.676507+00:00"},{"alias_kind":"pith_short_12","alias_value":"Q6CQSEQJDWBZ","created_at":"2026-07-05T11:33:02.676507+00:00"},{"alias_kind":"pith_short_16","alias_value":"Q6CQSEQJDWBZTW7A","created_at":"2026-07-05T11:33:02.676507+00:00"},{"alias_kind":"pith_short_8","alias_value":"Q6CQSEQJ","created_at":"2026-07-05T11:33:02.676507+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08733","citing_title":"Super Weights in LLMs and the Failure of Selective Training","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2606.08292","citing_title":"Necessary, Decodable and Reversible, Yet Not Transferable: A Stress Test for Attention-Head Role Claims","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03465","citing_title":"Rethinking the Role of Tensor Decompositions in Post-Training LLM Compression","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18898","citing_title":"A Two-Parameter Weibull Framework for Diagnosing Transformer Weight Distributions","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27401","citing_title":"Perturbation Probing: A Two-Pass-per-Prompt Diagnostic for FFN Behavioral Circuits in Aligned LLMs","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23475","citing_title":"Supernodes and Halos: Loss-Critical Hubs in LLM Feed-Forward Layers","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23036","citing_title":"Preserving Long-Tailed Expert Information in Mixture-of-Experts Tuning","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10098","citing_title":"Attention Sink in Transformers: A Survey on Utilization, Interpretation, and Mitigation","ref_index":189,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18231","citing_title":"AgenTEE: Confidential LLM Agent Execution on Edge Devices","ref_index":61,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J","json":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J.json","graph_json":"https://pith.science/api/pith-number/Q6CQSEQJDWBZTW7AXZUZIV7E5J/graph.json","events_json":"https://pith.science/api/pith-number/Q6CQSEQJDWBZTW7AXZUZIV7E5J/events.json","paper":"https://pith.science/paper/Q6CQSEQJ"},"agent_actions":{"view_html":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J","download_json":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J.json","view_paper":"https://pith.science/paper/Q6CQSEQJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.07191&json=true","fetch_graph":"https://pith.science/api/pith-number/Q6CQSEQJDWBZTW7AXZUZIV7E5J/graph.json","fetch_events":"https://pith.science/api/pith-number/Q6CQSEQJDWBZTW7AXZUZIV7E5J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J/action/storage_attestation","attest_author":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J/action/author_attestation","sign_citation":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J/action/citation_signature","submit_replication":"https://pith.science/pith/Q6CQSEQJDWBZTW7AXZUZIV7E5J/action/replication_record"}},"created_at":"2026-07-05T11:33:02.676507+00:00","updated_at":"2026-07-05T11:33:02.676507+00:00"}