{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WFKGNTFPJNKCDKGGRDQ645WA65","short_pith_number":"pith:WFKGNTFP","schema_version":"1.0","canonical_sha256":"b15466ccaf4b5421a8c688e1ee76c0f7639e25c775bd32a142b4504b6a510acb","source":{"kind":"arxiv","id":"2402.09656","version":4},"attestation_state":"computed","paper":{"title":"The Butterfly Effect of Model Editing: Few Edits Can Trigger Large Language Models Collapse","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Dawei Yin, Fei Sun, Wanli Yang, Xinyu Ma, Xueqi Cheng, Xun Liu","submitted_at":"2024-02-15T01:50:38Z","abstract_excerpt":"Although model editing has shown promise in revising knowledge in Large Language Models (LLMs), its impact on the inherent capabilities of LLMs is often overlooked. In this work, we reveal a critical phenomenon: even a single edit can trigger model collapse, manifesting as significant performance degradation in various benchmark tasks. However, benchmarking LLMs after each edit, while necessary to prevent such collapses, is impractically time-consuming and resource-intensive. To mitigate this, we propose using perplexity as a surrogate metric, validated by extensive experiments demonstrating c"},"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":"2402.09656","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2024-02-15T01:50:38Z","cross_cats_sorted":[],"title_canon_sha256":"b1f8a50c7f52c6e260b77f5c277a75c0ef4c9e2abdb73ad9610b4d7d3420996c","abstract_canon_sha256":"2d92ef0c7b6e9ae3f2340c273f216e2d136eed48af0e7f96f696483ea408c1a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:46.070492Z","signature_b64":"Iw/Y7Ra8qAmXNr+S9wgsOyLXxhNj6k2tDi55KLoahQ3JM7V491awYOna9+aMgCDxhyBJKetGeHFbMtp2Le1pAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b15466ccaf4b5421a8c688e1ee76c0f7639e25c775bd32a142b4504b6a510acb","last_reissued_at":"2026-07-05T08:27:46.069921Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:46.069921Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Butterfly Effect of Model Editing: Few Edits Can Trigger Large Language Models Collapse","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Dawei Yin, Fei Sun, Wanli Yang, Xinyu Ma, Xueqi Cheng, Xun Liu","submitted_at":"2024-02-15T01:50:38Z","abstract_excerpt":"Although model editing has shown promise in revising knowledge in Large Language Models (LLMs), its impact on the inherent capabilities of LLMs is often overlooked. In this work, we reveal a critical phenomenon: even a single edit can trigger model collapse, manifesting as significant performance degradation in various benchmark tasks. However, benchmarking LLMs after each edit, while necessary to prevent such collapses, is impractically time-consuming and resource-intensive. To mitigate this, we propose using perplexity as a surrogate metric, validated by extensive experiments demonstrating c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.09656","kind":"arxiv","version":4},"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/2402.09656/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":"2402.09656","created_at":"2026-07-05T08:27:46.069983+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.09656v4","created_at":"2026-07-05T08:27:46.069983+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.09656","created_at":"2026-07-05T08:27:46.069983+00:00"},{"alias_kind":"pith_short_12","alias_value":"WFKGNTFPJNKC","created_at":"2026-07-05T08:27:46.069983+00:00"},{"alias_kind":"pith_short_16","alias_value":"WFKGNTFPJNKCDKGG","created_at":"2026-07-05T08:27:46.069983+00:00"},{"alias_kind":"pith_short_8","alias_value":"WFKGNTFP","created_at":"2026-07-05T08:27:46.069983+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65","json":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65.json","graph_json":"https://pith.science/api/pith-number/WFKGNTFPJNKCDKGGRDQ645WA65/graph.json","events_json":"https://pith.science/api/pith-number/WFKGNTFPJNKCDKGGRDQ645WA65/events.json","paper":"https://pith.science/paper/WFKGNTFP"},"agent_actions":{"view_html":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65","download_json":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65.json","view_paper":"https://pith.science/paper/WFKGNTFP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.09656&json=true","fetch_graph":"https://pith.science/api/pith-number/WFKGNTFPJNKCDKGGRDQ645WA65/graph.json","fetch_events":"https://pith.science/api/pith-number/WFKGNTFPJNKCDKGGRDQ645WA65/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65/action/storage_attestation","attest_author":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65/action/author_attestation","sign_citation":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65/action/citation_signature","submit_replication":"https://pith.science/pith/WFKGNTFPJNKCDKGGRDQ645WA65/action/replication_record"}},"created_at":"2026-07-05T08:27:46.069983+00:00","updated_at":"2026-07-05T08:27:46.069983+00:00"}