{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:PCJG4TUC4EZN57IXVOQSJVFL6K","short_pith_number":"pith:PCJG4TUC","schema_version":"1.0","canonical_sha256":"78926e4e82e132defd17aba124d4abf2ab2843803b09df743a781e453f61d013","source":{"kind":"arxiv","id":"2305.14739","version":1},"attestation_state":"computed","paper":{"title":"Trusting Your Evidence: Hallucinate Less with Context-aware Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Luke Zettlemoyer, Mike Lewis, Scott Wen-tau Yih, Weijia Shi, Xiaochuang Han, Yulia Tsvetkov","submitted_at":"2023-05-24T05:19:15Z","abstract_excerpt":"Language models (LMs) often struggle to pay enough attention to the input context, and generate texts that are unfaithful or contain hallucinations. To mitigate this issue, we present context-aware decoding (CAD), which follows a contrastive output distribution that amplifies the difference between the output probabilities when a model is used with and without context. Our experiments show that CAD, without additional training, significantly improves the faithfulness of different LM families, including OPT, GPT, LLaMA and FLAN-T5 for summarization tasks (e.g., 14.3% gain for LLaMA in factualit"},"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":"2305.14739","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-24T05:19:15Z","cross_cats_sorted":[],"title_canon_sha256":"1bb10db10e5a351b36aa7f81ea154f71fdd7d6fbadcb116072d34e4e3a3dd1e9","abstract_canon_sha256":"7a02b984a9f90f37619454f0dd59f0e9fc26773d9128ffb30396f5a759f4407b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:13:27.423610Z","signature_b64":"mtQQWn305Aoliq9X61KSYQ1T7MW4w3poxiaOg3U6OIgM3hfH9ILnlBDmeYFqjlbWtJmpS5/J0o3ppoeE6pNMBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"78926e4e82e132defd17aba124d4abf2ab2843803b09df743a781e453f61d013","last_reissued_at":"2026-07-05T06:13:27.423144Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:13:27.423144Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Trusting Your Evidence: Hallucinate Less with Context-aware Decoding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Luke Zettlemoyer, Mike Lewis, Scott Wen-tau Yih, Weijia Shi, Xiaochuang Han, Yulia Tsvetkov","submitted_at":"2023-05-24T05:19:15Z","abstract_excerpt":"Language models (LMs) often struggle to pay enough attention to the input context, and generate texts that are unfaithful or contain hallucinations. To mitigate this issue, we present context-aware decoding (CAD), which follows a contrastive output distribution that amplifies the difference between the output probabilities when a model is used with and without context. Our experiments show that CAD, without additional training, significantly improves the faithfulness of different LM families, including OPT, GPT, LLaMA and FLAN-T5 for summarization tasks (e.g., 14.3% gain for LLaMA in factualit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.14739","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/2305.14739/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":"2305.14739","created_at":"2026-07-05T06:13:27.423200+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.14739v1","created_at":"2026-07-05T06:13:27.423200+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.14739","created_at":"2026-07-05T06:13:27.423200+00:00"},{"alias_kind":"pith_short_12","alias_value":"PCJG4TUC4EZN","created_at":"2026-07-05T06:13:27.423200+00:00"},{"alias_kind":"pith_short_16","alias_value":"PCJG4TUC4EZN57IX","created_at":"2026-07-05T06:13:27.423200+00:00"},{"alias_kind":"pith_short_8","alias_value":"PCJG4TUC","created_at":"2026-07-05T06:13:27.423200+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.21121","citing_title":"Answer Engineering: Local Trajectory Editing for Protocol-Constrained Decision Making in Large Language Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.25902","citing_title":"Reading the Finetuning Prior: Verbatim Content Recovery via Contrastive Decoding Diffing","ref_index":18,"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":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18792","citing_title":"Trust or Abstain? A Self-Aware RAG Approach","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2309.03883","citing_title":"DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models","ref_index":89,"is_internal_anchor":false},{"citing_arxiv_id":"2401.11817","citing_title":"Hallucination is Inevitable: An Innate Limitation of Large Language Models","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2311.05232","citing_title":"A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions","ref_index":288,"is_internal_anchor":false},{"citing_arxiv_id":"2309.01219","citing_title":"Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11141","citing_title":"Reducing Hallucination in Enterprise AI Workflows via Hybrid Utility Minimum Bayes Risk (HUMBR)","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16686","citing_title":"No-Worse Context-Aware Decoding: Preventing Neutral Regression in Context-Conditioned Generation","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K","json":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K.json","graph_json":"https://pith.science/api/pith-number/PCJG4TUC4EZN57IXVOQSJVFL6K/graph.json","events_json":"https://pith.science/api/pith-number/PCJG4TUC4EZN57IXVOQSJVFL6K/events.json","paper":"https://pith.science/paper/PCJG4TUC"},"agent_actions":{"view_html":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K","download_json":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K.json","view_paper":"https://pith.science/paper/PCJG4TUC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.14739&json=true","fetch_graph":"https://pith.science/api/pith-number/PCJG4TUC4EZN57IXVOQSJVFL6K/graph.json","fetch_events":"https://pith.science/api/pith-number/PCJG4TUC4EZN57IXVOQSJVFL6K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K/action/storage_attestation","attest_author":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K/action/author_attestation","sign_citation":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K/action/citation_signature","submit_replication":"https://pith.science/pith/PCJG4TUC4EZN57IXVOQSJVFL6K/action/replication_record"}},"created_at":"2026-07-05T06:13:27.423200+00:00","updated_at":"2026-07-05T06:13:27.423200+00:00"}