{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JS3MJM26KVFOSYMFQCH674OQPO","short_pith_number":"pith:JS3MJM26","schema_version":"1.0","canonical_sha256":"4cb6c4b35e554ae96185808feff1d07b9064972d4c7b0d814ddfd59ae3c36c8a","source":{"kind":"arxiv","id":"2410.16983","version":1},"attestation_state":"computed","paper":{"title":"Order Matters: Exploring Order Sensitivity in Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Tong Mo, Weiping Li, Xu Chu, Zhijie Tan","submitted_at":"2024-10-22T13:05:11Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) utilize multimodal contexts consisting of text, images, or videos to solve various multimodal tasks. However, we find that changing the order of multimodal input can cause the model's performance to fluctuate between advanced performance and random guessing. This phenomenon exists in both single-modality (text-only or image-only) and mixed-modality (image-text-pair) contexts. Furthermore, we demonstrate that popular MLLMs pay special attention to certain multimodal context positions, particularly the beginning and end. Leveraging this special attention,"},"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":"2410.16983","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-10-22T13:05:11Z","cross_cats_sorted":[],"title_canon_sha256":"cfad4979c3fa6fac752349f2e128693c5c608cd880626b198c48a832fb0ab522","abstract_canon_sha256":"16b425f12a615573dd7c23562cb8cbbcbdabaa2db67f357a045fad04cd325622"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:24:09.779645Z","signature_b64":"PcpMyiQfLI6b+i0p2pWqRy3w0dxcR6+Zv/A3/HxxwM7wA9OXqoof5nHBNOEgJ8TNEk/jI+8MczJ9YUWaNPjGAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4cb6c4b35e554ae96185808feff1d07b9064972d4c7b0d814ddfd59ae3c36c8a","last_reissued_at":"2026-07-05T09:24:09.779098Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:24:09.779098Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Order Matters: Exploring Order Sensitivity in Multimodal Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Tong Mo, Weiping Li, Xu Chu, Zhijie Tan","submitted_at":"2024-10-22T13:05:11Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) utilize multimodal contexts consisting of text, images, or videos to solve various multimodal tasks. However, we find that changing the order of multimodal input can cause the model's performance to fluctuate between advanced performance and random guessing. This phenomenon exists in both single-modality (text-only or image-only) and mixed-modality (image-text-pair) contexts. Furthermore, we demonstrate that popular MLLMs pay special attention to certain multimodal context positions, particularly the beginning and end. Leveraging this special attention,"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.16983","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/2410.16983/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":"2410.16983","created_at":"2026-07-05T09:24:09.779157+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.16983v1","created_at":"2026-07-05T09:24:09.779157+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.16983","created_at":"2026-07-05T09:24:09.779157+00:00"},{"alias_kind":"pith_short_12","alias_value":"JS3MJM26KVFO","created_at":"2026-07-05T09:24:09.779157+00:00"},{"alias_kind":"pith_short_16","alias_value":"JS3MJM26KVFOSYMF","created_at":"2026-07-05T09:24:09.779157+00:00"},{"alias_kind":"pith_short_8","alias_value":"JS3MJM26","created_at":"2026-07-05T09:24:09.779157+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26079","citing_title":"Same Evidence, Different Answer: Auditing Order Sensitivity in Multimodal Large Language Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09484","citing_title":"Detecting Differences Is Not Understanding Structure: Large Language Models Fail at Graph Isomorphism","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2606.16494","citing_title":"Lost at the End: Primacy Bias in Multimodal Retrieval-Augmented Question Answering","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19307","citing_title":"MetaRA: Metamorphic Robustness Assessment for Multimodal Large Language Model-based Visual Question Answering Systems","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11974","citing_title":"Towards Order Fairness: Mitigating LLMs Order Sensitivity through Dual Group Advantage Optimization","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11591","citing_title":"Logit-Attention Divergence: Mitigating Position Bias in Multi-Image Retrieval via Attention-Guided Calibration","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO","json":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO.json","graph_json":"https://pith.science/api/pith-number/JS3MJM26KVFOSYMFQCH674OQPO/graph.json","events_json":"https://pith.science/api/pith-number/JS3MJM26KVFOSYMFQCH674OQPO/events.json","paper":"https://pith.science/paper/JS3MJM26"},"agent_actions":{"view_html":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO","download_json":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO.json","view_paper":"https://pith.science/paper/JS3MJM26","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.16983&json=true","fetch_graph":"https://pith.science/api/pith-number/JS3MJM26KVFOSYMFQCH674OQPO/graph.json","fetch_events":"https://pith.science/api/pith-number/JS3MJM26KVFOSYMFQCH674OQPO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO/action/storage_attestation","attest_author":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO/action/author_attestation","sign_citation":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO/action/citation_signature","submit_replication":"https://pith.science/pith/JS3MJM26KVFOSYMFQCH674OQPO/action/replication_record"}},"created_at":"2026-07-05T09:24:09.779157+00:00","updated_at":"2026-07-05T09:24:09.779157+00:00"}