{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:Q7ENI4Q5JRIYF3N3TDRGU7JOYM","short_pith_number":"pith:Q7ENI4Q5","schema_version":"1.0","canonical_sha256":"87c8d4721d4c5182edbb98e26a7d2ec30678654ee6ab24d39257b8807adc3299","source":{"kind":"arxiv","id":"2405.20797","version":2},"attestation_state":"computed","paper":{"title":"Ovis: Structural Embedding Alignment for Multimodal Large Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Han-Jia Ye, Kaifu Zhang, Qing-Guo Chen, Shiyin Lu, Weihua Luo, Yang Li, Zhao Xu","submitted_at":"2024-05-31T13:59:18Z","abstract_excerpt":"Current Multimodal Large Language Models (MLLMs) typically integrate a pre-trained LLM with another pre-trained vision transformer through a connector, such as an MLP, endowing the LLM with visual capabilities. However, the misalignment between two embedding strategies in MLLMs -- the structural textual embeddings based on an embedding look-up table and the continuous embeddings generated directly by the vision encoder -- makes challenges for a more seamless fusion of visual and textual information. We propose Ovis, a novel MLLM architecture designed to structurally align visual and textual em"},"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":"2405.20797","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-05-31T13:59:18Z","cross_cats_sorted":["cs.AI","cs.CL","cs.LG"],"title_canon_sha256":"21f76fe3800bb4bd41c7941b8fdc9003bc361dab331f0c2a35d222d3ab4f081b","abstract_canon_sha256":"a4d94a4ba7dbac9fa5edd4970ce241f9a49811568e133384ee3893496ee54c01"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:32:41.519763Z","signature_b64":"nFUk37hta25uUcI6gfZ3Ok7crcfkzpPPDD52ygOe+QLmBb/7GYNXm6g8T6DnFJWu8VAqnNuQfOVvMD2c8aPcCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"87c8d4721d4c5182edbb98e26a7d2ec30678654ee6ab24d39257b8807adc3299","last_reissued_at":"2026-07-05T08:32:41.519287Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:32:41.519287Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Ovis: Structural Embedding Alignment for Multimodal Large Language Model","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Han-Jia Ye, Kaifu Zhang, Qing-Guo Chen, Shiyin Lu, Weihua Luo, Yang Li, Zhao Xu","submitted_at":"2024-05-31T13:59:18Z","abstract_excerpt":"Current Multimodal Large Language Models (MLLMs) typically integrate a pre-trained LLM with another pre-trained vision transformer through a connector, such as an MLP, endowing the LLM with visual capabilities. However, the misalignment between two embedding strategies in MLLMs -- the structural textual embeddings based on an embedding look-up table and the continuous embeddings generated directly by the vision encoder -- makes challenges for a more seamless fusion of visual and textual information. We propose Ovis, a novel MLLM architecture designed to structurally align visual and textual em"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.20797","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/2405.20797/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":"2405.20797","created_at":"2026-07-05T08:32:41.519349+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.20797v2","created_at":"2026-07-05T08:32:41.519349+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.20797","created_at":"2026-07-05T08:32:41.519349+00:00"},{"alias_kind":"pith_short_12","alias_value":"Q7ENI4Q5JRIY","created_at":"2026-07-05T08:32:41.519349+00:00"},{"alias_kind":"pith_short_16","alias_value":"Q7ENI4Q5JRIYF3N3","created_at":"2026-07-05T08:32:41.519349+00:00"},{"alias_kind":"pith_short_8","alias_value":"Q7ENI4Q5","created_at":"2026-07-05T08:32:41.519349+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":30,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25634","citing_title":"SSMNBench: Diagnosing Image-based Cross-View Human-Object Understanding via Single-View Sufficiency and Multi-View Necessity","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2606.22437","citing_title":"MMGist: A Comprehensive Multimodal Benchmark for 2027","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09132","citing_title":"Vision Language Model Helps Private Information De-Identification in Vision Data","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08034","citing_title":"Sci-Rho: A Multilingual Visually-Grounded Symbolic Benchmark for STEM Problems","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07801","citing_title":"Improving Multimodal Reasoning via Worst Dimension Optimization","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03264","citing_title":"PaddleOCR-VL-1.6: Expanding the Frontier of Document Parsing with Under-Optimized Region Refinement and Progressive Post-Training","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11651","citing_title":"Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31187","citing_title":"Learning to Deny: Action Denial in Multimodal Large Language Models","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2505.16416","citing_title":"Circle-RoPE: Cone-like Decoupled Rotary Positional Embedding for Large Vision-Language Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2602.04802","citing_title":"VISTA-Bench: Do Vision-Language Models Really Understand Visualized Text as Well as Pure Text?","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17823","citing_title":"Why We Look Where We Look: Emergent Human-like Fixations of a Foveated Visual Language Model Maximizing Scene Understanding","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11651","citing_title":"Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2506.09373","citing_title":"LPO: Towards Accurate GUI Agent Interaction via Location Preference Optimization","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2506.09522","citing_title":"Revisit What You See: Revealing Visual Semantics in Vision Tokens to Guide LVLM Decoding","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2501.01957","citing_title":"VITA-1.5: Towards GPT-4o Level Real-Time Vision and Speech Interaction","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2501.00321","citing_title":"OCRBench v2: An Improved Benchmark for Evaluating Large Multimodal Models on Visual Text Localization and Reasoning","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2512.10554","citing_title":"Grounding Everything in Tokens for Multimodal Large Language Models","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2512.14044","citing_title":"OmniDrive-R1: Reinforcement-driven Interleaved Multi-modal Chain-of-Thought for Trustworthy Vision-Language Autonomous Driving","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2411.10440","citing_title":"LLaVA-CoT: Let Vision Language Models Reason Step-by-Step","ref_index":39,"is_internal_anchor":false},{"citing_arxiv_id":"2602.00181","citing_title":"CamReasoner: Reinforcing Camera Movement Understanding via Structured Spatial Reasoning","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2602.08392","citing_title":"ST-BiBench: Benchmarking Multi-Stream Multimodal Coordination in Bimanual Embodied Tasks for MLLMs","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11651","citing_title":"Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11651","citing_title":"Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10576","citing_title":"SenseBench: A Benchmark for Remote Sensing Low-Level Visual Perception and Description in Large Vision-Language Models","ref_index":60,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25477","citing_title":"DDA-Thinker: Decoupled Dual-Atomic Reinforcement Learning for Reasoning-Driven Image Editing","ref_index":53,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM","json":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM.json","graph_json":"https://pith.science/api/pith-number/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/graph.json","events_json":"https://pith.science/api/pith-number/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/events.json","paper":"https://pith.science/paper/Q7ENI4Q5"},"agent_actions":{"view_html":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM","download_json":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM.json","view_paper":"https://pith.science/paper/Q7ENI4Q5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.20797&json=true","fetch_graph":"https://pith.science/api/pith-number/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/graph.json","fetch_events":"https://pith.science/api/pith-number/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/action/storage_attestation","attest_author":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/action/author_attestation","sign_citation":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/action/citation_signature","submit_replication":"https://pith.science/pith/Q7ENI4Q5JRIYF3N3TDRGU7JOYM/action/replication_record"}},"created_at":"2026-07-05T08:32:41.519349+00:00","updated_at":"2026-07-05T08:32:41.519349+00:00"}