{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BWVWFDX4BYB5N7VIJVYESBG6U5","short_pith_number":"pith:BWVWFDX4","schema_version":"1.0","canonical_sha256":"0dab628efc0e03d6fea84d704904dea76508a5272dc6a5ddfa679751bf95328d","source":{"kind":"arxiv","id":"2502.06415","version":2},"attestation_state":"computed","paper":{"title":"Systematic Outliers in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Jinqiao Wang, Ming Tang, Tao Yu, Xu Zhao, Yongqi An","submitted_at":"2025-02-10T12:54:17Z","abstract_excerpt":"Outliers have been widely observed in Large Language Models (LLMs), significantly impacting model performance and posing challenges for model compression. Understanding the functionality and formation mechanisms of these outliers is critically important. Existing works, however, largely focus on reducing the impact of outliers from an algorithmic perspective, lacking an in-depth investigation into their causes and roles. In this work, we provide a detailed analysis of the formation process, underlying causes, and functions of outliers in LLMs. We define and categorize three types of outliers-a"},"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":"2502.06415","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-10T12:54:17Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"c6e8e3a1ef70c9dd3a770262efe78cbc042e8f888dbbbfbebd0f84224da8659b","abstract_canon_sha256":"623276a1bb0497ff87419e9324ca8a92579c3a74eca5e1a7d1f59a033953b31a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:19:57.299624Z","signature_b64":"soQqg71+0dNylUjlCkg8jVRM8Rargg58+oYErU6PTG9c04hGBIUff/FLQfz/wJcH1yNUgbi3naKOxZsp/YcaCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0dab628efc0e03d6fea84d704904dea76508a5272dc6a5ddfa679751bf95328d","last_reissued_at":"2026-07-05T10:19:57.299155Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:19:57.299155Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Systematic Outliers in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Jinqiao Wang, Ming Tang, Tao Yu, Xu Zhao, Yongqi An","submitted_at":"2025-02-10T12:54:17Z","abstract_excerpt":"Outliers have been widely observed in Large Language Models (LLMs), significantly impacting model performance and posing challenges for model compression. Understanding the functionality and formation mechanisms of these outliers is critically important. Existing works, however, largely focus on reducing the impact of outliers from an algorithmic perspective, lacking an in-depth investigation into their causes and roles. In this work, we provide a detailed analysis of the formation process, underlying causes, and functions of outliers in LLMs. We define and categorize three types of outliers-a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.06415","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/2502.06415/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":"2502.06415","created_at":"2026-07-05T10:19:57.299212+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.06415v2","created_at":"2026-07-05T10:19:57.299212+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.06415","created_at":"2026-07-05T10:19:57.299212+00:00"},{"alias_kind":"pith_short_12","alias_value":"BWVWFDX4BYB5","created_at":"2026-07-05T10:19:57.299212+00:00"},{"alias_kind":"pith_short_16","alias_value":"BWVWFDX4BYB5N7VI","created_at":"2026-07-05T10:19:57.299212+00:00"},{"alias_kind":"pith_short_8","alias_value":"BWVWFDX4","created_at":"2026-07-05T10:19:57.299212+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08733","citing_title":"Super Weights in LLMs and the Failure of Selective Training","ref_index":2,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16147","citing_title":"Registers Matter for Pixel-Space Diffusion Transformers","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19660","citing_title":"OScaR: The Occam's Razor for Extreme KV Cache Quantization in LLMs and Beyond","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2512.02010","citing_title":"Four Over Six: More Accurate NVFP4 Quantization with Adaptive Block Scaling","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2601.14004","citing_title":"Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18231","citing_title":"AgenTEE: Confidential LLM Agent Execution on Edge Devices","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5","json":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5.json","graph_json":"https://pith.science/api/pith-number/BWVWFDX4BYB5N7VIJVYESBG6U5/graph.json","events_json":"https://pith.science/api/pith-number/BWVWFDX4BYB5N7VIJVYESBG6U5/events.json","paper":"https://pith.science/paper/BWVWFDX4"},"agent_actions":{"view_html":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5","download_json":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5.json","view_paper":"https://pith.science/paper/BWVWFDX4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.06415&json=true","fetch_graph":"https://pith.science/api/pith-number/BWVWFDX4BYB5N7VIJVYESBG6U5/graph.json","fetch_events":"https://pith.science/api/pith-number/BWVWFDX4BYB5N7VIJVYESBG6U5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5/action/storage_attestation","attest_author":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5/action/author_attestation","sign_citation":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5/action/citation_signature","submit_replication":"https://pith.science/pith/BWVWFDX4BYB5N7VIJVYESBG6U5/action/replication_record"}},"created_at":"2026-07-05T10:19:57.299212+00:00","updated_at":"2026-07-05T10:19:57.299212+00:00"}