{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KUXNBRMMRKZLZZSS74VTTPSXH6","short_pith_number":"pith:KUXNBRMM","schema_version":"1.0","canonical_sha256":"552ed0c58c8ab2bce652ff2b39be573f829cee0b062a44869556e5bbdb508fed","source":{"kind":"arxiv","id":"2404.04633","version":3},"attestation_state":"computed","paper":{"title":"Context versus Prior Knowledge in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aaron Schein, Jennifer C. White, Kevin Du, Niklas Stoehr, Ryan Cotterell, V\\'esteinn Sn{\\ae}bjarnarson","submitted_at":"2024-04-06T13:46:53Z","abstract_excerpt":"To answer a question, language models often need to integrate prior knowledge learned during pretraining and new information presented in context. We hypothesize that models perform this integration in a predictable way across different questions and contexts: models will rely more on prior knowledge for questions about entities (e.g., persons, places, etc.) that they are more familiar with due to higher exposure in the training corpus, and be more easily persuaded by some contexts than others. To formalize this problem, we propose two mutual information-based metrics to measure a model's depe"},"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":"2404.04633","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-04-06T13:46:53Z","cross_cats_sorted":[],"title_canon_sha256":"fe0eeb1a3589ab820c55190d2ea49a73ed5ecea064f00feb36064887870863a2","abstract_canon_sha256":"d5bda837e9f09556e28d1ccb81ab1edb959c90976c4d6c8fdff3b6ad1a4183ce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:32:39.336500Z","signature_b64":"diBGbGh2KdPsHCtH8b/FHxxuthCFzQIQcz6dMjykbkeMYbkNFbFkhOydk5Su7Itu8StM6CnRLd3/11I0nmBkAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"552ed0c58c8ab2bce652ff2b39be573f829cee0b062a44869556e5bbdb508fed","last_reissued_at":"2026-07-05T08:32:39.335940Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:32:39.335940Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Context versus Prior Knowledge in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aaron Schein, Jennifer C. White, Kevin Du, Niklas Stoehr, Ryan Cotterell, V\\'esteinn Sn{\\ae}bjarnarson","submitted_at":"2024-04-06T13:46:53Z","abstract_excerpt":"To answer a question, language models often need to integrate prior knowledge learned during pretraining and new information presented in context. We hypothesize that models perform this integration in a predictable way across different questions and contexts: models will rely more on prior knowledge for questions about entities (e.g., persons, places, etc.) that they are more familiar with due to higher exposure in the training corpus, and be more easily persuaded by some contexts than others. To formalize this problem, we propose two mutual information-based metrics to measure a model's depe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.04633","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/2404.04633/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":"2404.04633","created_at":"2026-07-05T08:32:39.336009+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.04633v3","created_at":"2026-07-05T08:32:39.336009+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.04633","created_at":"2026-07-05T08:32:39.336009+00:00"},{"alias_kind":"pith_short_12","alias_value":"KUXNBRMMRKZL","created_at":"2026-07-05T08:32:39.336009+00:00"},{"alias_kind":"pith_short_16","alias_value":"KUXNBRMMRKZLZZSS","created_at":"2026-07-05T08:32:39.336009+00:00"},{"alias_kind":"pith_short_8","alias_value":"KUXNBRMM","created_at":"2026-07-05T08:32:39.336009+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.00448","citing_title":"HERA: Improving Long Document Summarization using Large Language Models with Context Packaging and Reordering","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6","json":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6.json","graph_json":"https://pith.science/api/pith-number/KUXNBRMMRKZLZZSS74VTTPSXH6/graph.json","events_json":"https://pith.science/api/pith-number/KUXNBRMMRKZLZZSS74VTTPSXH6/events.json","paper":"https://pith.science/paper/KUXNBRMM"},"agent_actions":{"view_html":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6","download_json":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6.json","view_paper":"https://pith.science/paper/KUXNBRMM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.04633&json=true","fetch_graph":"https://pith.science/api/pith-number/KUXNBRMMRKZLZZSS74VTTPSXH6/graph.json","fetch_events":"https://pith.science/api/pith-number/KUXNBRMMRKZLZZSS74VTTPSXH6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6/action/storage_attestation","attest_author":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6/action/author_attestation","sign_citation":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6/action/citation_signature","submit_replication":"https://pith.science/pith/KUXNBRMMRKZLZZSS74VTTPSXH6/action/replication_record"}},"created_at":"2026-07-05T08:32:39.336009+00:00","updated_at":"2026-07-05T08:32:39.336009+00:00"}