{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JRBUWS5AGJXTP4RHNZOTMAMEQP","short_pith_number":"pith:JRBUWS5A","schema_version":"1.0","canonical_sha256":"4c434b4ba0326f37f2276e5d36018483c474b22b7b0930a930c567252cbfc643","source":{"kind":"arxiv","id":"2606.04928","version":1},"attestation_state":"computed","paper":{"title":"Data Attribution in Large Language Models via Bidirectional Gradient Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Fr\\'ed\\'eric Berdoz, Kaan Bayraktar, Luca A. Lanzend\\\"orfer, Roger Wattenhofer","submitted_at":"2026-06-03T14:21:53Z","abstract_excerpt":"Large Language Models (LLMs) are increasingly deployed across diverse applications, raising critical questions for governance, accountability, and data provenance. Understanding which training data most influenced a model's output remains a fundamental open problem. We address this challenge through training data attribution (TDA) for auto-regressive LLMs by expanding upon the inverse formulation: How would training data be affected if the model had seen the generated output during training? Our method perturbs the base model using bidirectional gradient optimization (gradient ascent and desce"},"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":"2606.04928","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-03T14:21:53Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6f45ae0a1aafece67619f89cc5876ce33cb1d4342804481955b41a3cbb24a5cb","abstract_canon_sha256":"a3a8b73954f0e77b0a18fd2aeb8d1c071d1a87f21bc1a9e1f200f9d4f2fda0a7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:09:56.163802Z","signature_b64":"PQRs1taKxiYR9zRPs4u1DdjbvED2ApBAO+RlHsZ3CP1hDwryFHboleFIUh/VeFE7H70cE+Lo15Cn/8x8L2NbDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4c434b4ba0326f37f2276e5d36018483c474b22b7b0930a930c567252cbfc643","last_reissued_at":"2026-06-04T01:09:56.162971Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:09:56.162971Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Data Attribution in Large Language Models via Bidirectional Gradient Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Fr\\'ed\\'eric Berdoz, Kaan Bayraktar, Luca A. Lanzend\\\"orfer, Roger Wattenhofer","submitted_at":"2026-06-03T14:21:53Z","abstract_excerpt":"Large Language Models (LLMs) are increasingly deployed across diverse applications, raising critical questions for governance, accountability, and data provenance. Understanding which training data most influenced a model's output remains a fundamental open problem. We address this challenge through training data attribution (TDA) for auto-regressive LLMs by expanding upon the inverse formulation: How would training data be affected if the model had seen the generated output during training? Our method perturbs the base model using bidirectional gradient optimization (gradient ascent and desce"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04928","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/2606.04928/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":"2606.04928","created_at":"2026-06-04T01:09:56.163101+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.04928v1","created_at":"2026-06-04T01:09:56.163101+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04928","created_at":"2026-06-04T01:09:56.163101+00:00"},{"alias_kind":"pith_short_12","alias_value":"JRBUWS5AGJXT","created_at":"2026-06-04T01:09:56.163101+00:00"},{"alias_kind":"pith_short_16","alias_value":"JRBUWS5AGJXTP4RH","created_at":"2026-06-04T01:09:56.163101+00:00"},{"alias_kind":"pith_short_8","alias_value":"JRBUWS5A","created_at":"2026-06-04T01:09:56.163101+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP","json":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP.json","graph_json":"https://pith.science/api/pith-number/JRBUWS5AGJXTP4RHNZOTMAMEQP/graph.json","events_json":"https://pith.science/api/pith-number/JRBUWS5AGJXTP4RHNZOTMAMEQP/events.json","paper":"https://pith.science/paper/JRBUWS5A"},"agent_actions":{"view_html":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP","download_json":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP.json","view_paper":"https://pith.science/paper/JRBUWS5A","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.04928&json=true","fetch_graph":"https://pith.science/api/pith-number/JRBUWS5AGJXTP4RHNZOTMAMEQP/graph.json","fetch_events":"https://pith.science/api/pith-number/JRBUWS5AGJXTP4RHNZOTMAMEQP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP/action/storage_attestation","attest_author":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP/action/author_attestation","sign_citation":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP/action/citation_signature","submit_replication":"https://pith.science/pith/JRBUWS5AGJXTP4RHNZOTMAMEQP/action/replication_record"}},"created_at":"2026-06-04T01:09:56.163101+00:00","updated_at":"2026-06-04T01:09:56.163101+00:00"}