{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:K4BRHDHC5ULJUITLODYXU67RPY","short_pith_number":"pith:K4BRHDHC","schema_version":"1.0","canonical_sha256":"5703138ce2ed169a226b70f17a7bf17e00883579e4140cfe21884768a4ed07f2","source":{"kind":"arxiv","id":"2309.15098","version":2},"attestation_state":"computed","paper":{"title":"Attention Satisfies: A Constraint-Satisfaction Lens on Factual Errors of Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Besmira Nushi, Ece Kamar, Erik Jones, Hamid Palangi, Mert Yuksekgonul, Ranjita Naik, Suriya Gunasekar, Varun Chandrasekaran","submitted_at":"2023-09-26T17:48:55Z","abstract_excerpt":"We investigate the internal behavior of Transformer-based Large Language Models (LLMs) when they generate factually incorrect text. We propose modeling factual queries as constraint satisfaction problems and use this framework to investigate how the LLM interacts internally with factual constraints. We find a strong positive relationship between the LLM's attention to constraint tokens and the factual accuracy of generations. We curate a suite of 10 datasets containing over 40,000 prompts to study the task of predicting factual errors with the Llama-2 family across all scales (7B, 13B, 70B). W"},"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":"2309.15098","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-09-26T17:48:55Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"d5c747ab7d8480b447ccacfcc1a3534ff00d03fb8c825ab7a0a29fc6dc7df3b1","abstract_canon_sha256":"794d54649c49fc404a495141272b84af5e83f89b72288d813b6d552cac916793"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:08:51.457287Z","signature_b64":"QR3eVSZ6LB2i3Jmv79Y5llqclN938ohxL2R8J0Ladxma+0QFcglc6/Iz0H4x+mM98s7AJKN4C5q8uj4aP80QCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5703138ce2ed169a226b70f17a7bf17e00883579e4140cfe21884768a4ed07f2","last_reissued_at":"2026-07-05T08:08:51.456796Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:08:51.456796Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attention Satisfies: A Constraint-Satisfaction Lens on Factual Errors of Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Besmira Nushi, Ece Kamar, Erik Jones, Hamid Palangi, Mert Yuksekgonul, Ranjita Naik, Suriya Gunasekar, Varun Chandrasekaran","submitted_at":"2023-09-26T17:48:55Z","abstract_excerpt":"We investigate the internal behavior of Transformer-based Large Language Models (LLMs) when they generate factually incorrect text. We propose modeling factual queries as constraint satisfaction problems and use this framework to investigate how the LLM interacts internally with factual constraints. We find a strong positive relationship between the LLM's attention to constraint tokens and the factual accuracy of generations. We curate a suite of 10 datasets containing over 40,000 prompts to study the task of predicting factual errors with the Llama-2 family across all scales (7B, 13B, 70B). W"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.15098","kind":"arxiv","version":2},"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/2309.15098/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":"2309.15098","created_at":"2026-07-05T08:08:51.456857+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.15098v2","created_at":"2026-07-05T08:08:51.456857+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.15098","created_at":"2026-07-05T08:08:51.456857+00:00"},{"alias_kind":"pith_short_12","alias_value":"K4BRHDHC5ULJ","created_at":"2026-07-05T08:08:51.456857+00:00"},{"alias_kind":"pith_short_16","alias_value":"K4BRHDHC5ULJUITL","created_at":"2026-07-05T08:08:51.456857+00:00"},{"alias_kind":"pith_short_8","alias_value":"K4BRHDHC","created_at":"2026-07-05T08:08:51.456857+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27786","citing_title":"SHIFT: Gate-Modulated Activation Steering for Knowledge Conflict Mitigation in Retrieval-Augmented Generation","ref_index":68,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY","json":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY.json","graph_json":"https://pith.science/api/pith-number/K4BRHDHC5ULJUITLODYXU67RPY/graph.json","events_json":"https://pith.science/api/pith-number/K4BRHDHC5ULJUITLODYXU67RPY/events.json","paper":"https://pith.science/paper/K4BRHDHC"},"agent_actions":{"view_html":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY","download_json":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY.json","view_paper":"https://pith.science/paper/K4BRHDHC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.15098&json=true","fetch_graph":"https://pith.science/api/pith-number/K4BRHDHC5ULJUITLODYXU67RPY/graph.json","fetch_events":"https://pith.science/api/pith-number/K4BRHDHC5ULJUITLODYXU67RPY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY/action/storage_attestation","attest_author":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY/action/author_attestation","sign_citation":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY/action/citation_signature","submit_replication":"https://pith.science/pith/K4BRHDHC5ULJUITLODYXU67RPY/action/replication_record"}},"created_at":"2026-07-05T08:08:51.456857+00:00","updated_at":"2026-07-05T08:08:51.456857+00:00"}