{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:5HNSSHDZYUWVSPLSQX67GDVPC5","short_pith_number":"pith:5HNSSHDZ","schema_version":"1.0","canonical_sha256":"e9db291c79c52d593d7285fdf30eaf177b759d000dcdbecf2aae177c5a177cb0","source":{"kind":"arxiv","id":"2002.08484","version":3},"attestation_state":"computed","paper":{"title":"Estimating Training Data Influence by Tracing Gradient Descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Frederick Liu, Garima Pruthi, Mukund Sundararajan, Satyen Kale","submitted_at":"2020-02-19T22:40:32Z","abstract_excerpt":"We introduce a method called TracIn that computes the influence of a training example on a prediction made by the model. The idea is to trace how the loss on the test point changes during the training process whenever the training example of interest was utilized. We provide a scalable implementation of TracIn via: (a) a first-order gradient approximation to the exact computation, (b) saved checkpoints of standard training procedures, and (c) cherry-picking layers of a deep neural network. In contrast with previously proposed methods, TracIn is simple to implement; all it needs is the ability "},"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":"2002.08484","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-02-19T22:40:32Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"add8883c2541101fd9325a212d6e684c542a279cfa2d78aaf2ec27e788decbe3","abstract_canon_sha256":"cb1b5ca7b3ee9ff4228ad39a1b3a93c345c8f70b52eeaaac697dfb7dc2c47551"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:51:39.635845Z","signature_b64":"eHllJqXf2HbENciMXZduFgycwbMeduneXZRST46fi+2QgX1/WtrL7yppNlY82JD1i7Et8HzGOPXmLnL8vQTODw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e9db291c79c52d593d7285fdf30eaf177b759d000dcdbecf2aae177c5a177cb0","last_reissued_at":"2026-07-05T01:51:39.635387Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:51:39.635387Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Estimating Training Data Influence by Tracing Gradient Descent","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Frederick Liu, Garima Pruthi, Mukund Sundararajan, Satyen Kale","submitted_at":"2020-02-19T22:40:32Z","abstract_excerpt":"We introduce a method called TracIn that computes the influence of a training example on a prediction made by the model. The idea is to trace how the loss on the test point changes during the training process whenever the training example of interest was utilized. We provide a scalable implementation of TracIn via: (a) a first-order gradient approximation to the exact computation, (b) saved checkpoints of standard training procedures, and (c) cherry-picking layers of a deep neural network. In contrast with previously proposed methods, TracIn is simple to implement; all it needs is the ability "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2002.08484","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/2002.08484/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":"2002.08484","created_at":"2026-07-05T01:51:39.635444+00:00"},{"alias_kind":"arxiv_version","alias_value":"2002.08484v3","created_at":"2026-07-05T01:51:39.635444+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2002.08484","created_at":"2026-07-05T01:51:39.635444+00:00"},{"alias_kind":"pith_short_12","alias_value":"5HNSSHDZYUWV","created_at":"2026-07-05T01:51:39.635444+00:00"},{"alias_kind":"pith_short_16","alias_value":"5HNSSHDZYUWVSPLS","created_at":"2026-07-05T01:51:39.635444+00:00"},{"alias_kind":"pith_short_8","alias_value":"5HNSSHDZ","created_at":"2026-07-05T01:51:39.635444+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21306","citing_title":"Towards Dys-XAI: Influence-Based Explanations for Dysarthria Severity Assessment","ref_index":37,"is_internal_anchor":false},{"citing_arxiv_id":"2606.05800","citing_title":"SALT: When More Rollouts Don't Help in Group-Based Policy Optimization and How to Make Them Matter","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2502.00270","citing_title":"DUET: Optimizing Training Data Mixtures via Feedback from Unseen Evaluation Tasks","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09404","citing_title":"Let the Target Select for Itself: Data Selection via Target-Aligned Paths","ref_index":35,"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":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5","json":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5.json","graph_json":"https://pith.science/api/pith-number/5HNSSHDZYUWVSPLSQX67GDVPC5/graph.json","events_json":"https://pith.science/api/pith-number/5HNSSHDZYUWVSPLSQX67GDVPC5/events.json","paper":"https://pith.science/paper/5HNSSHDZ"},"agent_actions":{"view_html":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5","download_json":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5.json","view_paper":"https://pith.science/paper/5HNSSHDZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2002.08484&json=true","fetch_graph":"https://pith.science/api/pith-number/5HNSSHDZYUWVSPLSQX67GDVPC5/graph.json","fetch_events":"https://pith.science/api/pith-number/5HNSSHDZYUWVSPLSQX67GDVPC5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5/action/storage_attestation","attest_author":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5/action/author_attestation","sign_citation":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5/action/citation_signature","submit_replication":"https://pith.science/pith/5HNSSHDZYUWVSPLSQX67GDVPC5/action/replication_record"}},"created_at":"2026-07-05T01:51:39.635444+00:00","updated_at":"2026-07-05T01:51:39.635444+00:00"}