{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:MQ7D3HTD22I6DRJZXL25IM6JIB","short_pith_number":"pith:MQ7D3HTD","schema_version":"1.0","canonical_sha256":"643e3d9e63d691e1c539baf5d433c9406a7e9a80a09c234b63546d4458339fcd","source":{"kind":"arxiv","id":"2409.19998","version":2},"attestation_state":"computed","paper":{"title":"Do Influence Functions Work on Large Language Models?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jun Sun, Wei Zhao, Yige Li, Zhe Li","submitted_at":"2024-09-30T06:50:18Z","abstract_excerpt":"Influence functions are important for quantifying the impact of individual training data points on a model's predictions. Although extensive research has been conducted on influence functions in traditional machine learning models, their application to large language models (LLMs) has been limited. In this work, we conduct a systematic study to address a key question: do influence functions work on LLMs? Specifically, we evaluate influence functions across multiple tasks and find that they consistently perform poorly in most settings. Our further investigation reveals that their poor performan"},"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":"2409.19998","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-09-30T06:50:18Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"dd22d02404db60ec16090d333d9c76e7f768eeb43dbeb76671c6344b9913a6cf","abstract_canon_sha256":"cb2f6702c4a690623eab5b79cce81b968a67e4d69f8180464c40ca4785327c3e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:52:16.550325Z","signature_b64":"25DfOXblv4Aeq8Ds5h+ICO6n41uhpuFJJ82Ev8vIhVEmiq014+vKhUOdGW0Xv0lsirwAEI8iQ1h+k6/O+z1oCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"643e3d9e63d691e1c539baf5d433c9406a7e9a80a09c234b63546d4458339fcd","last_reissued_at":"2026-07-05T09:52:16.548940Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:52:16.548940Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do Influence Functions Work on Large Language Models?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Jun Sun, Wei Zhao, Yige Li, Zhe Li","submitted_at":"2024-09-30T06:50:18Z","abstract_excerpt":"Influence functions are important for quantifying the impact of individual training data points on a model's predictions. Although extensive research has been conducted on influence functions in traditional machine learning models, their application to large language models (LLMs) has been limited. In this work, we conduct a systematic study to address a key question: do influence functions work on LLMs? Specifically, we evaluate influence functions across multiple tasks and find that they consistently perform poorly in most settings. Our further investigation reveals that their poor performan"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.19998","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/2409.19998/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":"2409.19998","created_at":"2026-07-05T09:52:16.548998+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.19998v2","created_at":"2026-07-05T09:52:16.548998+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.19998","created_at":"2026-07-05T09:52:16.548998+00:00"},{"alias_kind":"pith_short_12","alias_value":"MQ7D3HTD22I6","created_at":"2026-07-05T09:52:16.548998+00:00"},{"alias_kind":"pith_short_16","alias_value":"MQ7D3HTD22I6DRJZ","created_at":"2026-07-05T09:52:16.548998+00:00"},{"alias_kind":"pith_short_8","alias_value":"MQ7D3HTD","created_at":"2026-07-05T09:52:16.548998+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.18307","citing_title":"DRIFT: Refining Instruction Data via On-Policy Data Attribution","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05029","citing_title":"Validity Threats for Foundation Model Research","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2606.23591","citing_title":"Quantifying the Agreement Between Data-Influence and Data-Similarity to Understand LLM Behavior","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2602.10995","citing_title":"A Human-Centric Framework for Data Attribution in Large Language Models","ref_index":113,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14194","citing_title":"GradShield: Alignment Preserving Finetuning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07769","citing_title":"An Empirical Study on Influence-Based Pretraining Data Selection for Code Large Language Models","ref_index":25,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB","json":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB.json","graph_json":"https://pith.science/api/pith-number/MQ7D3HTD22I6DRJZXL25IM6JIB/graph.json","events_json":"https://pith.science/api/pith-number/MQ7D3HTD22I6DRJZXL25IM6JIB/events.json","paper":"https://pith.science/paper/MQ7D3HTD"},"agent_actions":{"view_html":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB","download_json":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB.json","view_paper":"https://pith.science/paper/MQ7D3HTD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.19998&json=true","fetch_graph":"https://pith.science/api/pith-number/MQ7D3HTD22I6DRJZXL25IM6JIB/graph.json","fetch_events":"https://pith.science/api/pith-number/MQ7D3HTD22I6DRJZXL25IM6JIB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB/action/storage_attestation","attest_author":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB/action/author_attestation","sign_citation":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB/action/citation_signature","submit_replication":"https://pith.science/pith/MQ7D3HTD22I6DRJZXL25IM6JIB/action/replication_record"}},"created_at":"2026-07-05T09:52:16.548998+00:00","updated_at":"2026-07-05T09:52:16.548998+00:00"}