Research map / Vision-language and multimodal / Cross-modal learning
Knowledge Completes the Vision: A Multimodal Entity-aware Retrieval-Augmented Generation Framework for News Image Captioning
Xiaoxing You, Qiang Huang, Lingyu Li, Chi Zhang, Xiaopeng Liu, Min Zhang, Jun Yu
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 |
|---|---|---|---|
| MERGE over EAMA | EAMA · GoodNews | CIDEr: 87.70 -> 94.54 | Improved |
| MERGE over EAMA | EAMA · NYTimes800k | CIDEr: 87.00 -> 88.16 | Improved |
| MERGE over Zhou et al. on unseen Visual News | Zhou et al. 2022 · Visual News | CIDEr: 107.60 -> 127.77 | Improved |
| MERGE vs Xu et al. 2024a on precision | Xu et al. 2024a · NYTimes800k | NER precision: 32.38 -> 31.87 | Hurt |
| fine-tune InstructBLIP (ablation) | InstructBLIP · GoodNews | CIDEr: 24.42 -> 84.8 | Improved |
| add HCMA Stage 1 (ablation) | InstructBLIP w/ FT · GoodNews | CIDEr: 84.8 -> 84.83 | No clear effect |
| add HCMA Stage 3 global summary (ablation) | HCMA Stage 1+2 · GoodNews | CIDEr: 85.24 -> 86.08 | Improved |
| add HCMA Stage 3 global summary (ablation) | HCMA Stage 1+2 · NYTimes800k | CIDEr: 76.81 -> 80.14 | Improved |
| add HCMA Stage 2 sentence selection (ablation) | HCMA Stage 1 · NYTimes800k | NER F1: 30.82 -> 30.10 | Hurt |
| add RAS 1 entity matching (ablation) | InstructBLIP w/ FT · GoodNews | CIDEr: 84.8 -> 91.52 | Improved |
| add RAS 2 background KG alone (ablation) | InstructBLIP w/ FT · NYTimes800k | CIDEr: 73.68 -> 74.43 | No clear effect |
| add RAS 2 to RAS 1 (ablation) | RMKI RAS 1 · Visual News | CIDEr: 116.84 -> 123.50 | Improved |
Closest papers by meaning
- A Dual-Way Enhanced Framework from Text Matching Point of View for Multimodal Entity Linking (AAAI 2024 · similarity 0.87)
- ARK: A Dual-Axis Multimodal Retrieval Benchmark along Reasoning and Knowledge (NeurIPS 2026 · similarity 0.86)
- Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval (NeurIPS 2025 · similarity 0.86)
- VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning (CVPR 2024 · similarity 0.86)
- Patch Matters: Training-free Fine-grained Image Caption Enhancement via Local Perception (CVPR 2025 · similarity 0.86)
- TIGeR: Unifying Text-to-Image Generation and Retrieval with Large Multimodal Models (ICLR 2025 · similarity 0.85)
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.
