{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RQRI2VG66MF6RRK7BDSWPXE2P3","short_pith_number":"pith:RQRI2VG6","schema_version":"1.0","canonical_sha256":"8c228d54def30be8c55f08e567dc9a7ef0404f35380630efab553fa739c68c56","source":{"kind":"arxiv","id":"2409.17357","version":1},"attestation_state":"computed","paper":{"title":"Revisiting inverse Hessian vector products for calculating influence functions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Yang Liu, Yegor Klochkov","submitted_at":"2024-09-25T21:05:35Z","abstract_excerpt":"Influence functions are a popular tool for attributing a model's output to training data. The traditional approach relies on the calculation of inverse Hessian-vector products (iHVP), but the classical solver \"Linear time Stochastic Second-order Algorithm\" (LiSSA, Agarwal et al. (2017)) is often deemed impractical for large models due to expensive computation and hyperparameter tuning. We show that the three hyperparameters -- the scaling factor, the batch size, and the number of steps -- can be chosen depending on the spectral properties of the Hessian, particularly its trace and largest eige"},"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":"2409.17357","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-09-25T21:05:35Z","cross_cats_sorted":[],"title_canon_sha256":"9e323099da8ee5e2bd8645885f53b84f9bf1bf31af724151c1d43d97abcd5a2c","abstract_canon_sha256":"d003218495dc88c47aaf1057046a0a49612b28d7c3cbf6a390a5de17c3bd1640"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:11:52.160114Z","signature_b64":"Iyq008FMx8PWxHRaTOYPVK5KiewxWjCT55Dv0aqmG3ve1eQFPxqA0m2KAQykhpM0XtRRlK50ATK21dXdurr2CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8c228d54def30be8c55f08e567dc9a7ef0404f35380630efab553fa739c68c56","last_reissued_at":"2026-07-05T09:11:52.159615Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:11:52.159615Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Revisiting inverse Hessian vector products for calculating influence functions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Yang Liu, Yegor Klochkov","submitted_at":"2024-09-25T21:05:35Z","abstract_excerpt":"Influence functions are a popular tool for attributing a model's output to training data. The traditional approach relies on the calculation of inverse Hessian-vector products (iHVP), but the classical solver \"Linear time Stochastic Second-order Algorithm\" (LiSSA, Agarwal et al. (2017)) is often deemed impractical for large models due to expensive computation and hyperparameter tuning. We show that the three hyperparameters -- the scaling factor, the batch size, and the number of steps -- can be chosen depending on the spectral properties of the Hessian, particularly its trace and largest eige"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2409.17357","kind":"arxiv","version":1},"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/2409.17357/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":"2409.17357","created_at":"2026-07-05T09:11:52.159677+00:00"},{"alias_kind":"arxiv_version","alias_value":"2409.17357v1","created_at":"2026-07-05T09:11:52.159677+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2409.17357","created_at":"2026-07-05T09:11:52.159677+00:00"},{"alias_kind":"pith_short_12","alias_value":"RQRI2VG66MF6","created_at":"2026-07-05T09:11:52.159677+00:00"},{"alias_kind":"pith_short_16","alias_value":"RQRI2VG66MF6RRK7","created_at":"2026-07-05T09:11:52.159677+00:00"},{"alias_kind":"pith_short_8","alias_value":"RQRI2VG6","created_at":"2026-07-05T09:11:52.159677+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.00510","citing_title":"Prototype Language Models","ref_index":169,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3","json":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3.json","graph_json":"https://pith.science/api/pith-number/RQRI2VG66MF6RRK7BDSWPXE2P3/graph.json","events_json":"https://pith.science/api/pith-number/RQRI2VG66MF6RRK7BDSWPXE2P3/events.json","paper":"https://pith.science/paper/RQRI2VG6"},"agent_actions":{"view_html":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3","download_json":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3.json","view_paper":"https://pith.science/paper/RQRI2VG6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2409.17357&json=true","fetch_graph":"https://pith.science/api/pith-number/RQRI2VG66MF6RRK7BDSWPXE2P3/graph.json","fetch_events":"https://pith.science/api/pith-number/RQRI2VG66MF6RRK7BDSWPXE2P3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3/action/storage_attestation","attest_author":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3/action/author_attestation","sign_citation":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3/action/citation_signature","submit_replication":"https://pith.science/pith/RQRI2VG66MF6RRK7BDSWPXE2P3/action/replication_record"}},"created_at":"2026-07-05T09:11:52.159677+00:00","updated_at":"2026-07-05T09:11:52.159677+00:00"}