{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:Z3FGY6C4MIJAGSCCRALBRYBEN5","short_pith_number":"pith:Z3FGY6C4","schema_version":"1.0","canonical_sha256":"ceca6c785c6212034842881618e0246f77b47b74e30b5f0a8f98ea0e2651bcda","source":{"kind":"arxiv","id":"1703.04730","version":3},"attestation_state":"computed","paper":{"title":"Understanding Black-box Predictions via Influence Functions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"stat.ML","authors_text":"Pang Wei Koh, Percy Liang","submitted_at":"2017-03-14T21:07:01Z","abstract_excerpt":"How can we explain the predictions of a black-box model? In this paper, we use influence functions -- a classic technique from robust statistics -- to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, 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":"1703.04730","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2017-03-14T21:07:01Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"447236ef78c42d76f63bc3df91f6786059e6a26a84e67c0195230f73e7596eb2","abstract_canon_sha256":"012ecbe800d021179405589d66fbacfd57fdc1ac17ac8b13b307983fb6bf4c68"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:02:48.582654Z","signature_b64":"/4eTyExXFptGZdNnX4f79NXYZWBVx8oNsHIk6oiLu/d4VnhIcfyBNktnTdtv/KuY1+3xVP0pqJiQffOoj0v3BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ceca6c785c6212034842881618e0246f77b47b74e30b5f0a8f98ea0e2651bcda","last_reissued_at":"2026-07-05T02:02:48.582187Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:02:48.582187Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding Black-box Predictions via Influence Functions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"stat.ML","authors_text":"Pang Wei Koh, Percy Liang","submitted_at":"2017-03-14T21:07:01Z","abstract_excerpt":"How can we explain the predictions of a black-box model? In this paper, we use influence functions -- a classic technique from robust statistics -- to trace a model's prediction through the learning algorithm and back to its training data, thereby identifying training points most responsible for a given prediction. To scale up influence functions to modern machine learning settings, we develop a simple, efficient implementation that requires only oracle access to gradients and Hessian-vector products. We show that even on non-convex and non-differentiable models where the theory breaks down, a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1703.04730","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/1703.04730/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":"1703.04730","created_at":"2026-07-05T02:02:48.582244+00:00"},{"alias_kind":"arxiv_version","alias_value":"1703.04730v3","created_at":"2026-07-05T02:02:48.582244+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1703.04730","created_at":"2026-07-05T02:02:48.582244+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z3FGY6C4MIJA","created_at":"2026-07-05T02:02:48.582244+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z3FGY6C4MIJAGSCC","created_at":"2026-07-05T02:02:48.582244+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z3FGY6C4","created_at":"2026-07-05T02:02:48.582244+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21306","citing_title":"Towards Dys-XAI: Influence-Based Explanations for Dysarthria Severity Assessment","ref_index":38,"is_internal_anchor":false},{"citing_arxiv_id":"2607.00325","citing_title":"Watermarking for Proprietary Dataset Protection","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21422","citing_title":"PRISM: Preference-Aware Influence Function Based Data Selection Method for Efficient Fine-Tuning","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26971","citing_title":"RLVR Datasets and Where to Find Them: Tracing Data Lineage for Better Training Data","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30462","citing_title":"idSCD: Identifying Training Datasets through Semantic Correlation Descriptors","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"1907.08657","citing_title":"Learning More From Less: Towards Strengthening Weak Supervision for Ad-Hoc Retrieval","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"1909.01066","citing_title":"Language Models as Knowledge Bases?","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.28154","citing_title":"Mapping data sensitivities in global QCD analysis with linear response and influence functions","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08007","citing_title":"Interpreting Reinforcement Learning Agents with Susceptibilities","ref_index":119,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16197","citing_title":"Sketching the Readout of Large Language Models for Scalable Data Attribution and Valuation","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5","json":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5.json","graph_json":"https://pith.science/api/pith-number/Z3FGY6C4MIJAGSCCRALBRYBEN5/graph.json","events_json":"https://pith.science/api/pith-number/Z3FGY6C4MIJAGSCCRALBRYBEN5/events.json","paper":"https://pith.science/paper/Z3FGY6C4"},"agent_actions":{"view_html":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5","download_json":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5.json","view_paper":"https://pith.science/paper/Z3FGY6C4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1703.04730&json=true","fetch_graph":"https://pith.science/api/pith-number/Z3FGY6C4MIJAGSCCRALBRYBEN5/graph.json","fetch_events":"https://pith.science/api/pith-number/Z3FGY6C4MIJAGSCCRALBRYBEN5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5/action/storage_attestation","attest_author":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5/action/author_attestation","sign_citation":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5/action/citation_signature","submit_replication":"https://pith.science/pith/Z3FGY6C4MIJAGSCCRALBRYBEN5/action/replication_record"}},"created_at":"2026-07-05T02:02:48.582244+00:00","updated_at":"2026-07-05T02:02:48.582244+00:00"}