{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:PGHHDTZJ7FW3JO5LLFAZFMZO6C","short_pith_number":"pith:PGHHDTZJ","schema_version":"1.0","canonical_sha256":"798e71cf29f96db4bbab594192b32ef0b36368b8841c30ae79ac5832f118b6d5","source":{"kind":"arxiv","id":"2406.01920","version":1},"attestation_state":"computed","paper":{"title":"CODE: Contrasting Self-generated Description to Combat Hallucination in Large Multi-modal Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Hyunjun Kim, Junho Kim, Yeonju Kim, Yong Man Ro","submitted_at":"2024-06-04T03:04:21Z","abstract_excerpt":"Large Multi-modal Models (LMMs) have recently demonstrated remarkable abilities in visual context understanding and coherent response generation. However, alongside these advancements, the issue of hallucinations has emerged as a significant challenge, producing erroneous responses that are unrelated to the visual contents. In this paper, we introduce a novel contrastive-based decoding method, COuntering DEscription Contrastive Decoding (CODE), which leverages self-generated descriptions as contrasting references during the decoding phase of LMMs to address hallucination issues. CODE utilizes "},"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":"2406.01920","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-06-04T03:04:21Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"13dd80bfcd713b6091b484b0291df1673f2131296d775fc83fc4732ea88e1235","abstract_canon_sha256":"0b71914682faca10acf1d7c21a43f764947f6e5435a3f8337e5dd0a06c06c759"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:11.678236Z","signature_b64":"fbFPVL1fvwhIChMV1ZYsyytPt4Yxnb6oKyOEocR5wux/TOQ+oam0TWuyzhSJ/6Bd53UpUjEFm2VQx8QVtFR3Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"798e71cf29f96db4bbab594192b32ef0b36368b8841c30ae79ac5832f118b6d5","last_reissued_at":"2026-07-05T08:27:11.677661Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:11.677661Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"CODE: Contrasting Self-generated Description to Combat Hallucination in Large Multi-modal Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Hyunjun Kim, Junho Kim, Yeonju Kim, Yong Man Ro","submitted_at":"2024-06-04T03:04:21Z","abstract_excerpt":"Large Multi-modal Models (LMMs) have recently demonstrated remarkable abilities in visual context understanding and coherent response generation. However, alongside these advancements, the issue of hallucinations has emerged as a significant challenge, producing erroneous responses that are unrelated to the visual contents. In this paper, we introduce a novel contrastive-based decoding method, COuntering DEscription Contrastive Decoding (CODE), which leverages self-generated descriptions as contrasting references during the decoding phase of LMMs to address hallucination issues. CODE utilizes "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.01920","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/2406.01920/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":"2406.01920","created_at":"2026-07-05T08:27:11.677732+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.01920v1","created_at":"2026-07-05T08:27:11.677732+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.01920","created_at":"2026-07-05T08:27:11.677732+00:00"},{"alias_kind":"pith_short_12","alias_value":"PGHHDTZJ7FW3","created_at":"2026-07-05T08:27:11.677732+00:00"},{"alias_kind":"pith_short_16","alias_value":"PGHHDTZJ7FW3JO5L","created_at":"2026-07-05T08:27:11.677732+00:00"},{"alias_kind":"pith_short_8","alias_value":"PGHHDTZJ","created_at":"2026-07-05T08:27:11.677732+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.27993","citing_title":"Rethinking Visual Neglect: Steering via Context-Preference for MLLM Hallucination Mitigation","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21027","citing_title":"HypEHR: Hyperbolic Modeling of Electronic Health Records for Efficient Question Answering","ref_index":265,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C","json":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C.json","graph_json":"https://pith.science/api/pith-number/PGHHDTZJ7FW3JO5LLFAZFMZO6C/graph.json","events_json":"https://pith.science/api/pith-number/PGHHDTZJ7FW3JO5LLFAZFMZO6C/events.json","paper":"https://pith.science/paper/PGHHDTZJ"},"agent_actions":{"view_html":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C","download_json":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C.json","view_paper":"https://pith.science/paper/PGHHDTZJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.01920&json=true","fetch_graph":"https://pith.science/api/pith-number/PGHHDTZJ7FW3JO5LLFAZFMZO6C/graph.json","fetch_events":"https://pith.science/api/pith-number/PGHHDTZJ7FW3JO5LLFAZFMZO6C/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C/action/storage_attestation","attest_author":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C/action/author_attestation","sign_citation":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C/action/citation_signature","submit_replication":"https://pith.science/pith/PGHHDTZJ7FW3JO5LLFAZFMZO6C/action/replication_record"}},"created_at":"2026-07-05T08:27:11.677732+00:00","updated_at":"2026-07-05T08:27:11.677732+00:00"}