{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:27KSWOPANMCBFD2T34TSY7P2JD","short_pith_number":"pith:27KSWOPA","schema_version":"1.0","canonical_sha256":"d7d52b39e06b04128f53df272c7dfa48f39dd8f97a0bd6f53b74396afa2f5ac3","source":{"kind":"arxiv","id":"2507.12455","version":3},"attestation_state":"computed","paper":{"title":"Mitigating Object Hallucinations via Sentence-Level Early Intervention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Li Jiang, Senqiao Yang, Shangpin Peng, Zhuotao Tian","submitted_at":"2025-07-16T17:55:43Z","abstract_excerpt":"Multimodal large language models (MLLMs) have revolutionized cross-modal understanding but continue to struggle with hallucinations - fabricated content contradicting visual inputs. Existing hallucination mitigation methods either incur prohibitive computational costs or introduce distribution mismatches between training data and model outputs. We identify a critical insight: hallucinations predominantly emerge at the early stages of text generation and propagate through subsequent outputs. To address this, we propose SENTINEL (Sentence-level Early iNtervention Through IN-domain prEference Lea"},"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":"2507.12455","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-07-16T17:55:43Z","cross_cats_sorted":[],"title_canon_sha256":"dfd70f1bb248ec9f43eb7c6f8b8bfd769285d51d0d7395f2ad6a1217118468bc","abstract_canon_sha256":"78f47c326cd14c2f52ed25b5b48c46d71cdedf91b6602adcd9af76f675e38bb2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:01:05.641956Z","signature_b64":"QfpGsIeHiQJZkaTqdic5AugtJDCGXT6b376kdI7YNnSCp2Ut0pqau2pVMnEMMwJH3P3nRAHG82g1J8r/XzFRDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d7d52b39e06b04128f53df272c7dfa48f39dd8f97a0bd6f53b74396afa2f5ac3","last_reissued_at":"2026-05-25T02:01:05.641134Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:01:05.641134Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating Object Hallucinations via Sentence-Level Early Intervention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Li Jiang, Senqiao Yang, Shangpin Peng, Zhuotao Tian","submitted_at":"2025-07-16T17:55:43Z","abstract_excerpt":"Multimodal large language models (MLLMs) have revolutionized cross-modal understanding but continue to struggle with hallucinations - fabricated content contradicting visual inputs. Existing hallucination mitigation methods either incur prohibitive computational costs or introduce distribution mismatches between training data and model outputs. We identify a critical insight: hallucinations predominantly emerge at the early stages of text generation and propagate through subsequent outputs. To address this, we propose SENTINEL (Sentence-level Early iNtervention Through IN-domain prEference Lea"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.12455","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/2507.12455/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":"2507.12455","created_at":"2026-05-25T02:01:05.641210+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.12455v3","created_at":"2026-05-25T02:01:05.641210+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.12455","created_at":"2026-05-25T02:01:05.641210+00:00"},{"alias_kind":"pith_short_12","alias_value":"27KSWOPANMCB","created_at":"2026-05-25T02:01:05.641210+00:00"},{"alias_kind":"pith_short_16","alias_value":"27KSWOPANMCBFD2T","created_at":"2026-05-25T02:01:05.641210+00:00"},{"alias_kind":"pith_short_8","alias_value":"27KSWOPA","created_at":"2026-05-25T02:01:05.641210+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2604.17982","citing_title":"Mitigating Multimodal Hallucination via Phase-wise Self-reward","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2604.21027","citing_title":"HypEHR: Hyperbolic Modeling of Electronic Health Records for Efficient Question Answering","ref_index":222,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD","json":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD.json","graph_json":"https://pith.science/api/pith-number/27KSWOPANMCBFD2T34TSY7P2JD/graph.json","events_json":"https://pith.science/api/pith-number/27KSWOPANMCBFD2T34TSY7P2JD/events.json","paper":"https://pith.science/paper/27KSWOPA"},"agent_actions":{"view_html":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD","download_json":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD.json","view_paper":"https://pith.science/paper/27KSWOPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.12455&json=true","fetch_graph":"https://pith.science/api/pith-number/27KSWOPANMCBFD2T34TSY7P2JD/graph.json","fetch_events":"https://pith.science/api/pith-number/27KSWOPANMCBFD2T34TSY7P2JD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD/action/storage_attestation","attest_author":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD/action/author_attestation","sign_citation":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD/action/citation_signature","submit_replication":"https://pith.science/pith/27KSWOPANMCBFD2T34TSY7P2JD/action/replication_record"}},"created_at":"2026-05-25T02:01:05.641210+00:00","updated_at":"2026-05-25T02:01:05.641210+00:00"}