{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:R4MHWONRQR6WOMAAZZK6RRYUP4","short_pith_number":"pith:R4MHWONR","schema_version":"1.0","canonical_sha256":"8f187b39b1847d673000ce55e8c7147f3508205ff3f9e12a19f85dadf9505077","source":{"kind":"arxiv","id":"2404.01019","version":3},"attestation_state":"computed","paper":{"title":"Source-Aware Training Enables Knowledge Attribution in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Wadden, Emma Strubell, Hao Peng, Honglak Lee, Iz Beltagy, Lu Wang, Muhammad Khalifa","submitted_at":"2024-04-01T09:39:38Z","abstract_excerpt":"Large language models (LLMs) learn a vast amount of knowledge during pretraining, but they are often oblivious to the source(s) of such knowledge. We investigate the problem of intrinsic source citation, where LLMs are required to cite the pretraining source supporting a generated response. Intrinsic source citation can enhance LLM transparency, interpretability, and verifiability. To give LLMs such ability, we explore source-aware training -- a recipe that involves (i) training the LLM to associate unique source document identifiers with the knowledge in each document, followed by (ii) an ins"},"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":"2404.01019","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-01T09:39:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"2dfad183cb860d5b2e096ea6dc68d565081a02bbf668f2daa6c0c3a16615d5c3","abstract_canon_sha256":"8dde0da238f7408cc9f56bad6c4e872246e3176dc1a955110043f7962661aedf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:54:48.830027Z","signature_b64":"hw4XUAMK5NGP38xkNzd6H3CBzR5RuTjwOqkXB69PGJKELntvKwkkS5bK6OqLlgdGokV6XuuGnjLze/SN41FUAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8f187b39b1847d673000ce55e8c7147f3508205ff3f9e12a19f85dadf9505077","last_reissued_at":"2026-07-05T08:54:48.829476Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:54:48.829476Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Source-Aware Training Enables Knowledge Attribution in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"David Wadden, Emma Strubell, Hao Peng, Honglak Lee, Iz Beltagy, Lu Wang, Muhammad Khalifa","submitted_at":"2024-04-01T09:39:38Z","abstract_excerpt":"Large language models (LLMs) learn a vast amount of knowledge during pretraining, but they are often oblivious to the source(s) of such knowledge. We investigate the problem of intrinsic source citation, where LLMs are required to cite the pretraining source supporting a generated response. Intrinsic source citation can enhance LLM transparency, interpretability, and verifiability. To give LLMs such ability, we explore source-aware training -- a recipe that involves (i) training the LLM to associate unique source document identifiers with the knowledge in each document, followed by (ii) an ins"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.01019","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/2404.01019/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":"2404.01019","created_at":"2026-07-05T08:54:48.829539+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.01019v3","created_at":"2026-07-05T08:54:48.829539+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.01019","created_at":"2026-07-05T08:54:48.829539+00:00"},{"alias_kind":"pith_short_12","alias_value":"R4MHWONRQR6W","created_at":"2026-07-05T08:54:48.829539+00:00"},{"alias_kind":"pith_short_16","alias_value":"R4MHWONRQR6WOMAA","created_at":"2026-07-05T08:54:48.829539+00:00"},{"alias_kind":"pith_short_8","alias_value":"R4MHWONR","created_at":"2026-07-05T08:54:48.829539+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.01284","citing_title":"Chain of Evidence: Pixel-Level Visual Attribution for Iterative Retrieval-Augmented Generation","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2409.10102","citing_title":"Trustworthiness in Retrieval-Augmented Generation Systems: A Survey","ref_index":136,"is_internal_anchor":false},{"citing_arxiv_id":"2511.21613","citing_title":"Beyond URLs: Metadata Diversity and Position for Efficient LLM Pretraining","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05687","citing_title":"DataDignity: Training Data Attribution for Large Language Models","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01284","citing_title":"Chain of Evidence: Pixel-Level Visual Attribution for Iterative Retrieval-Augmented Generation","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4","json":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4.json","graph_json":"https://pith.science/api/pith-number/R4MHWONRQR6WOMAAZZK6RRYUP4/graph.json","events_json":"https://pith.science/api/pith-number/R4MHWONRQR6WOMAAZZK6RRYUP4/events.json","paper":"https://pith.science/paper/R4MHWONR"},"agent_actions":{"view_html":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4","download_json":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4.json","view_paper":"https://pith.science/paper/R4MHWONR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.01019&json=true","fetch_graph":"https://pith.science/api/pith-number/R4MHWONRQR6WOMAAZZK6RRYUP4/graph.json","fetch_events":"https://pith.science/api/pith-number/R4MHWONRQR6WOMAAZZK6RRYUP4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4/action/storage_attestation","attest_author":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4/action/author_attestation","sign_citation":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4/action/citation_signature","submit_replication":"https://pith.science/pith/R4MHWONRQR6WOMAAZZK6RRYUP4/action/replication_record"}},"created_at":"2026-07-05T08:54:48.829539+00:00","updated_at":"2026-07-05T08:54:48.829539+00:00"}