{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:2FGB4TT5SZKOUWGFB6JTOMNRQQ","short_pith_number":"pith:2FGB4TT5","schema_version":"1.0","canonical_sha256":"d14c1e4e7d9654ea58c50f933731b184004166c4ef6f2bbc939279aa01e77f0c","source":{"kind":"arxiv","id":"2010.05873","version":1},"attestation_state":"computed","paper":{"title":"Controlled Hallucinations: Learning to Generate Faithfully from Noisy Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Katja Filippova","submitted_at":"2020-10-12T17:25:02Z","abstract_excerpt":"Neural text generation (data- or text-to-text) demonstrates remarkable performance when training data is abundant which for many applications is not the case. To collect a large corpus of parallel data, heuristic rules are often used but they inevitably let noise into the data, such as phrases in the output which cannot be explained by the input. Consequently, models pick up on the noise and may hallucinate--generate fluent but unsupported text. Our contribution is a simple but powerful technique to treat such hallucinations as a controllable aspect of the generated text, without dismissing an"},"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":"2010.05873","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-10-12T17:25:02Z","cross_cats_sorted":[],"title_canon_sha256":"dff6876f1c90360aed00f4acf84125364073c37f3c50bc9c2e93c8b87bc0f395","abstract_canon_sha256":"76f55031a5374d4bb3cad8e7b8e3684095e4c9a6163923773c110f35e9cc2547"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:42:11.308128Z","signature_b64":"IUQQm0SBuNVBzhhWQWJcNPQqxjZK7nqpv586AwUzYbBa4AGFune8RsFoU2/YK+2hSf+iB3UPqKR+4eI0sOQuCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d14c1e4e7d9654ea58c50f933731b184004166c4ef6f2bbc939279aa01e77f0c","last_reissued_at":"2026-07-05T01:42:11.307673Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:42:11.307673Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Controlled Hallucinations: Learning to Generate Faithfully from Noisy Data","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Katja Filippova","submitted_at":"2020-10-12T17:25:02Z","abstract_excerpt":"Neural text generation (data- or text-to-text) demonstrates remarkable performance when training data is abundant which for many applications is not the case. To collect a large corpus of parallel data, heuristic rules are often used but they inevitably let noise into the data, such as phrases in the output which cannot be explained by the input. Consequently, models pick up on the noise and may hallucinate--generate fluent but unsupported text. Our contribution is a simple but powerful technique to treat such hallucinations as a controllable aspect of the generated text, without dismissing an"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.05873","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/2010.05873/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":"2010.05873","created_at":"2026-07-05T01:42:11.307731+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.05873v1","created_at":"2026-07-05T01:42:11.307731+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.05873","created_at":"2026-07-05T01:42:11.307731+00:00"},{"alias_kind":"pith_short_12","alias_value":"2FGB4TT5SZKO","created_at":"2026-07-05T01:42:11.307731+00:00"},{"alias_kind":"pith_short_16","alias_value":"2FGB4TT5SZKOUWGF","created_at":"2026-07-05T01:42:11.307731+00:00"},{"alias_kind":"pith_short_8","alias_value":"2FGB4TT5","created_at":"2026-07-05T01:42:11.307731+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2308.05374","citing_title":"Trustworthy LLMs: a Survey and Guideline for Evaluating Large Language Models' Alignment","ref_index":82,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ","json":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ.json","graph_json":"https://pith.science/api/pith-number/2FGB4TT5SZKOUWGFB6JTOMNRQQ/graph.json","events_json":"https://pith.science/api/pith-number/2FGB4TT5SZKOUWGFB6JTOMNRQQ/events.json","paper":"https://pith.science/paper/2FGB4TT5"},"agent_actions":{"view_html":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ","download_json":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ.json","view_paper":"https://pith.science/paper/2FGB4TT5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.05873&json=true","fetch_graph":"https://pith.science/api/pith-number/2FGB4TT5SZKOUWGFB6JTOMNRQQ/graph.json","fetch_events":"https://pith.science/api/pith-number/2FGB4TT5SZKOUWGFB6JTOMNRQQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ/action/storage_attestation","attest_author":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ/action/author_attestation","sign_citation":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ/action/citation_signature","submit_replication":"https://pith.science/pith/2FGB4TT5SZKOUWGFB6JTOMNRQQ/action/replication_record"}},"created_at":"2026-07-05T01:42:11.307731+00:00","updated_at":"2026-07-05T01:42:11.307731+00:00"}