{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:KDQ4D6MFNLESKM4P3GPKZSCDFG","short_pith_number":"pith:KDQ4D6MF","schema_version":"1.0","canonical_sha256":"50e1c1f9856ac925338fd99eacc8432987a48658fc28aba5fb3c10b2f74ea4e1","source":{"kind":"arxiv","id":"2310.03149","version":4},"attestation_state":"computed","paper":{"title":"Attributing Learned Concepts in Neural Networks to Training Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Charles Godfrey, Davis Brown, Henry Kvinge, Jonathan Tu, Madelyn Shapiro, Nicholas Konz","submitted_at":"2023-10-04T20:26:59Z","abstract_excerpt":"By now there is substantial evidence that deep learning models learn certain human-interpretable features as part of their internal representations of data. As having the right (or wrong) concepts is critical to trustworthy machine learning systems, it is natural to ask which inputs from the model's original training set were most important for learning a concept at a given layer. To answer this, we combine data attribution methods with methods for probing the concepts learned by a model. Training network and probe ensembles for two concept datasets on a range of network layers, we use the rec"},"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":"2310.03149","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-10-04T20:26:59Z","cross_cats_sorted":["cs.AI","cs.CV"],"title_canon_sha256":"b805e86de8d44459c77f531a0b5f835521d9a7e80166069f689dd7b3b5a22800","abstract_canon_sha256":"23852c0df6e6ebc97a74408f60865d19d36249f9ae7de5af025c216f905a7f71"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:28:30.271946Z","signature_b64":"5AQX3KSF6zBGIhnuj9YuRd4iZFKlf8MXTsLPiRSJF1jToXtAtO4+Gb19WXpgpvTg+441w/aFBNLdry9jGJDSCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"50e1c1f9856ac925338fd99eacc8432987a48658fc28aba5fb3c10b2f74ea4e1","last_reissued_at":"2026-07-05T07:28:30.271505Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:28:30.271505Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attributing Learned Concepts in Neural Networks to Training Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CV"],"primary_cat":"cs.LG","authors_text":"Charles Godfrey, Davis Brown, Henry Kvinge, Jonathan Tu, Madelyn Shapiro, Nicholas Konz","submitted_at":"2023-10-04T20:26:59Z","abstract_excerpt":"By now there is substantial evidence that deep learning models learn certain human-interpretable features as part of their internal representations of data. As having the right (or wrong) concepts is critical to trustworthy machine learning systems, it is natural to ask which inputs from the model's original training set were most important for learning a concept at a given layer. To answer this, we combine data attribution methods with methods for probing the concepts learned by a model. Training network and probe ensembles for two concept datasets on a range of network layers, we use the rec"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.03149","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/2310.03149/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":"2310.03149","created_at":"2026-07-05T07:28:30.271565+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.03149v4","created_at":"2026-07-05T07:28:30.271565+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.03149","created_at":"2026-07-05T07:28:30.271565+00:00"},{"alias_kind":"pith_short_12","alias_value":"KDQ4D6MFNLES","created_at":"2026-07-05T07:28:30.271565+00:00"},{"alias_kind":"pith_short_16","alias_value":"KDQ4D6MFNLESKM4P","created_at":"2026-07-05T07:28:30.271565+00:00"},{"alias_kind":"pith_short_8","alias_value":"KDQ4D6MF","created_at":"2026-07-05T07:28:30.271565+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.05508","citing_title":"Rigorous Interpretation Is a Form of Evaluation","ref_index":79,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG","json":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG.json","graph_json":"https://pith.science/api/pith-number/KDQ4D6MFNLESKM4P3GPKZSCDFG/graph.json","events_json":"https://pith.science/api/pith-number/KDQ4D6MFNLESKM4P3GPKZSCDFG/events.json","paper":"https://pith.science/paper/KDQ4D6MF"},"agent_actions":{"view_html":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG","download_json":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG.json","view_paper":"https://pith.science/paper/KDQ4D6MF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.03149&json=true","fetch_graph":"https://pith.science/api/pith-number/KDQ4D6MFNLESKM4P3GPKZSCDFG/graph.json","fetch_events":"https://pith.science/api/pith-number/KDQ4D6MFNLESKM4P3GPKZSCDFG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG/action/storage_attestation","attest_author":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG/action/author_attestation","sign_citation":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG/action/citation_signature","submit_replication":"https://pith.science/pith/KDQ4D6MFNLESKM4P3GPKZSCDFG/action/replication_record"}},"created_at":"2026-07-05T07:28:30.271565+00:00","updated_at":"2026-07-05T07:28:30.271565+00:00"}