Research map / Vision-language and multimodal / Human and embodied modeling
Open Visual Knowledge Extraction via Relation-Oriented Multimodality Model Prompting
Hejie Cui, Xinyu Fang, Zihan Zhang, Ran Xu, Xuan Kan, Xin Liu, Yue Yu, Manling Li, Yangqiu Song, Carl Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use relation-oriented region prompting and format-free multimodal generation | OpenVik · Open visual knowledge generation | BLEU: 0.275 -> 0.280 | Improved |
| Use relation-oriented region prompting and format-free multimodal generation | OpenVik · Open visual knowledge generation | Validity: 0.897 -> 0.907 | Improved |
| Use relation-oriented region prompting and format-free multimodal generation | OpenVik · Open visual knowledge generation | Conformity: 0.871 -> 0.883 | Improved |
| Use relation-oriented region prompting and format-free multimodal generation | OpenVik · Open visual knowledge generation | Freshness: 0.795 -> 0.809 | Improved |
| Use relation-oriented region prompting and format-free multimodal generation | OpenVik · Open visual knowledge generation | Diversity: 0.547 -> 0.619 | Improved |
| Enrich captions with OpenVik relational context | Zero-shot CLIP retrieval · MS-COCO text-to-image retrieval | Avg: more than 6.0% | Improved |
| Enrich event descriptions with OpenVik verb-related contexts | Zero-shot CLIP and BLIP · Grounded Situation Recognition | Accuracy | Improved |
| Add two-level entity and relation context from OpenVik | Zero-shot VCR baseline · Visual Commonsense Reasoning | Accuracy: above 3.0% | Improved |
| Add the inter-sequence information variety regularizer (ablation) | OpenVik visual knowledge generator · Generated knowledge quality | Diversity | Improved |
| Add diversity-driven random dropping of low-quality, frequent descriptions (ablation) | OpenVik training data · Generated knowledge quality | Diversity | Improved |
| Augment entities and relations with external knowledge resources (ablation) | OpenVik training data · Generated knowledge quality | Freshness | Improved |
| Augment entities and relations with external knowledge resources (ablation) | OpenVik training data · Generated knowledge quality | Diversity | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- BLIP2 — BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (ICML 2022) · better on 1 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MMKE-Bench: A Multimodal Editing Benchmark for Diverse Visual Knowledge (ICLR 2025 · similarity 0.88)
- LION: Empowering Multimodal Large Language Model with Dual-Level Visual Knowledge (CVPR 2024 · similarity 0.87)
- MMFineReason: Closing the Multimodal Reasoning Gap via Open Data-Centric Methods (NeurIPS 2026 · similarity 0.87)
- UKnow: A Unified Knowledge Protocol with Multimodal Knowledge Graph Datasets for Reasoning and Vision-Language Pre-Training (NeurIPS 2024 · similarity 0.87)
- VLM-R³: Region Recognition, Reasoning, and Refinement for Enhanced Multimodal Chain-of-Thought (NeurIPS 2025 · similarity 0.87)
- TVI-CoT: Text-Visual Interleaved Chain-of-Thought Reasoning for Multimodal Understanding (ICML 2026 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
