{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:WHAAMI2JZTV3GQNREPN5USZBGU","short_pith_number":"pith:WHAAMI2J","schema_version":"1.0","canonical_sha256":"b1c0062349ccebb341b123dbda4b213538e32c5e6379bddad18d1c354cf3be2d","source":{"kind":"arxiv","id":"2407.08039","version":1},"attestation_state":"computed","paper":{"title":"Knowledge Overshadowing Causes Amalgamated Hallucination in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Heng Ji, Jiateng Liu, Jing Li, Manling Li, Pengfei Yu, Sha Li, Yi R. Fung, Yuji Zhang","submitted_at":"2024-07-10T20:37:42Z","abstract_excerpt":"Hallucination is often regarded as a major impediment for using large language models (LLMs), especially for knowledge-intensive tasks. Even when the training corpus consists solely of true statements, language models still generate hallucinations in the form of amalgamations of multiple facts. We coin this phenomenon as ``knowledge overshadowing'': when we query knowledge from a language model with multiple conditions, some conditions overshadow others, leading to hallucinated outputs. This phenomenon partially stems from training data imbalance, which we verify on both pretrained models and "},"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":"2407.08039","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-07-10T20:37:42Z","cross_cats_sorted":[],"title_canon_sha256":"dcd09f57c7cdc873e3b53ce8495e99b0220ce73935c10049e80c3cd24b59a1a8","abstract_canon_sha256":"e1c4359eed13596b4c1ea2c418b920331359d183cee4c819954e8476713dd327"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:42:42.206078Z","signature_b64":"v38QyQKVviXROQfthekrV9rOpA0vGfuEuTEtNym4qguiBYwT0YikQSS/w2L/7eJ5fW+EbV8i0c4z0WAbf5E0CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b1c0062349ccebb341b123dbda4b213538e32c5e6379bddad18d1c354cf3be2d","last_reissued_at":"2026-07-05T08:42:42.205580Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:42:42.205580Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Knowledge Overshadowing Causes Amalgamated Hallucination in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Heng Ji, Jiateng Liu, Jing Li, Manling Li, Pengfei Yu, Sha Li, Yi R. Fung, Yuji Zhang","submitted_at":"2024-07-10T20:37:42Z","abstract_excerpt":"Hallucination is often regarded as a major impediment for using large language models (LLMs), especially for knowledge-intensive tasks. Even when the training corpus consists solely of true statements, language models still generate hallucinations in the form of amalgamations of multiple facts. We coin this phenomenon as ``knowledge overshadowing'': when we query knowledge from a language model with multiple conditions, some conditions overshadow others, leading to hallucinated outputs. This phenomenon partially stems from training data imbalance, which we verify on both pretrained models and "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.08039","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/2407.08039/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":"2407.08039","created_at":"2026-07-05T08:42:42.205648+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.08039v1","created_at":"2026-07-05T08:42:42.205648+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.08039","created_at":"2026-07-05T08:42:42.205648+00:00"},{"alias_kind":"pith_short_12","alias_value":"WHAAMI2JZTV3","created_at":"2026-07-05T08:42:42.205648+00:00"},{"alias_kind":"pith_short_16","alias_value":"WHAAMI2JZTV3GQNR","created_at":"2026-07-05T08:42:42.205648+00:00"},{"alias_kind":"pith_short_8","alias_value":"WHAAMI2J","created_at":"2026-07-05T08:42:42.205648+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.02443","citing_title":"HalluScan: A Systematic Benchmark for Detecting and Mitigating Hallucinations in Instruction-Following LLMs","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2412.20760","citing_title":"Attributing Culture-Conditioned Generations to Pretraining Corpora","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02443","citing_title":"HalluScan: A Systematic Benchmark for Detecting and Mitigating Hallucinations in Instruction-Following LLMs","ref_index":46,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU","json":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU.json","graph_json":"https://pith.science/api/pith-number/WHAAMI2JZTV3GQNREPN5USZBGU/graph.json","events_json":"https://pith.science/api/pith-number/WHAAMI2JZTV3GQNREPN5USZBGU/events.json","paper":"https://pith.science/paper/WHAAMI2J"},"agent_actions":{"view_html":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU","download_json":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU.json","view_paper":"https://pith.science/paper/WHAAMI2J","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.08039&json=true","fetch_graph":"https://pith.science/api/pith-number/WHAAMI2JZTV3GQNREPN5USZBGU/graph.json","fetch_events":"https://pith.science/api/pith-number/WHAAMI2JZTV3GQNREPN5USZBGU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU/action/storage_attestation","attest_author":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU/action/author_attestation","sign_citation":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU/action/citation_signature","submit_replication":"https://pith.science/pith/WHAAMI2JZTV3GQNREPN5USZBGU/action/replication_record"}},"created_at":"2026-07-05T08:42:42.205648+00:00","updated_at":"2026-07-05T08:42:42.205648+00:00"}