{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:K63ALBBTEIWU66AZG2FAXTW5H6","short_pith_number":"pith:K63ALBBT","schema_version":"1.0","canonical_sha256":"57b6058433222d4f7819368a0bcedd3fad87addb6f73864922ea7f2fa5418099","source":{"kind":"arxiv","id":"2503.07772","version":2},"attestation_state":"computed","paper":{"title":"Hallucinatory Image Tokens: A Training-free EAZY Approach on Detecting and Mitigating Object Hallucinations in LVLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Jing Jia, Liwei Che, Ruixiang Tang, Tony Qingze Liu, Vladimir Pavlovic, Weiyi Qin","submitted_at":"2025-03-10T18:53:39Z","abstract_excerpt":"Despite their remarkable potential, Large Vision-Language Models (LVLMs) still face challenges with object hallucination, a problem where their generated outputs mistakenly incorporate objects that do not actually exist. Although most works focus on addressing this issue within the language-model backbone, our work shifts the focus to the image input source, investigating how specific image tokens contribute to hallucinations. Our analysis reveals a striking finding: a small subset of image tokens with high attention scores are the primary drivers of object hallucination. By removing these hal"},"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":"2503.07772","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-03-10T18:53:39Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b0164ab71a9c84e69e479dd6ced5fe9dc3d2cd3e0e2448386f386603187e4f07","abstract_canon_sha256":"c0f52e60b09cbf37e8bd42bd5f10522a8c2e5d01d7382a9f35fd5ff12cec5529"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:32:25.861587Z","signature_b64":"0pKM6/5/Txg/JA0vjRM3JEhXrGLE94frfn37xGb+Oc+6611KomNktfUGi0auHrw1Gtb1mUyMBPSl7ZK3HK4rBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"57b6058433222d4f7819368a0bcedd3fad87addb6f73864922ea7f2fa5418099","last_reissued_at":"2026-07-05T11:32:25.861087Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:32:25.861087Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hallucinatory Image Tokens: A Training-free EAZY Approach on Detecting and Mitigating Object Hallucinations in LVLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Jing Jia, Liwei Che, Ruixiang Tang, Tony Qingze Liu, Vladimir Pavlovic, Weiyi Qin","submitted_at":"2025-03-10T18:53:39Z","abstract_excerpt":"Despite their remarkable potential, Large Vision-Language Models (LVLMs) still face challenges with object hallucination, a problem where their generated outputs mistakenly incorporate objects that do not actually exist. Although most works focus on addressing this issue within the language-model backbone, our work shifts the focus to the image input source, investigating how specific image tokens contribute to hallucinations. Our analysis reveals a striking finding: a small subset of image tokens with high attention scores are the primary drivers of object hallucination. By removing these hal"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.07772","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/2503.07772/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":"2503.07772","created_at":"2026-07-05T11:32:25.861148+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.07772v2","created_at":"2026-07-05T11:32:25.861148+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.07772","created_at":"2026-07-05T11:32:25.861148+00:00"},{"alias_kind":"pith_short_12","alias_value":"K63ALBBTEIWU","created_at":"2026-07-05T11:32:25.861148+00:00"},{"alias_kind":"pith_short_16","alias_value":"K63ALBBTEIWU66AZ","created_at":"2026-07-05T11:32:25.861148+00:00"},{"alias_kind":"pith_short_8","alias_value":"K63ALBBT","created_at":"2026-07-05T11:32:25.861148+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27596","citing_title":"Dismantling Pathological Shortcuts: A Causal Framework for Faithful LVLM Decoding","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2404.18930","citing_title":"Hallucination of Multimodal Large Language Models: A Survey","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17982","citing_title":"Mitigating Multimodal Hallucination via Phase-wise Self-reward","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21027","citing_title":"HypEHR: Hyperbolic Modeling of Electronic Health Records for Efficient Question Answering","ref_index":220,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6","json":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6.json","graph_json":"https://pith.science/api/pith-number/K63ALBBTEIWU66AZG2FAXTW5H6/graph.json","events_json":"https://pith.science/api/pith-number/K63ALBBTEIWU66AZG2FAXTW5H6/events.json","paper":"https://pith.science/paper/K63ALBBT"},"agent_actions":{"view_html":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6","download_json":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6.json","view_paper":"https://pith.science/paper/K63ALBBT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.07772&json=true","fetch_graph":"https://pith.science/api/pith-number/K63ALBBTEIWU66AZG2FAXTW5H6/graph.json","fetch_events":"https://pith.science/api/pith-number/K63ALBBTEIWU66AZG2FAXTW5H6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6/action/storage_attestation","attest_author":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6/action/author_attestation","sign_citation":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6/action/citation_signature","submit_replication":"https://pith.science/pith/K63ALBBTEIWU66AZG2FAXTW5H6/action/replication_record"}},"created_at":"2026-07-05T11:32:25.861148+00:00","updated_at":"2026-07-05T11:32:25.861148+00:00"}