Research map / Vision-language and multimodal / Visual question answering
REVEAL: Retrieval-Augmented Visual-Language Pre-Training With Multi-Source Multimodal Knowledge Memory
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 REVEAL instead of ReVIVE (Ensemble) | OK-VQA · OK-VQA | accuracy (%): 58.0 -> 59.1 | Helped |
| Use REVEAL instead of GPV-2 | A-OKVQA · A-OKVQA | accuracy (%): +3.6% | Helped |
| Use REVEAL instead of CoCa | Image captioning · MSCOCO Karpathy-test | CIDEr: 143.6 -> 145.4 | Helped |
| Use REVEAL instead of CoCa | Image captioning · NoCaps val | CIDEr: 122.4 -> 123.0 | Helped |
| Add CC12M to the full knowledge-source set (ablation) | REVEAL-Base · OK-VQA validation | accuracy (%): 1.3% | Helped |
| Add VQA-v2 to the full knowledge-source set (ablation) | REVEAL-Base · OK-VQA validation | accuracy (%): 0.6% | Helped |
| Add WikiData to the full knowledge-source set (ablation) | REVEAL-Base · OK-VQA validation | accuracy (%): 1.1% | Helped |
| Use WIT as the only knowledge source (ablation) | REVEAL-Base · OK-VQA validation | accuracy (%): 53.1 | Mixed |
| Use attentive fusion instead of Perplexity Distill (ablation) | REVEAL-Base retriever training · WIT retrieval | Acc@10: 0.704 -> 0.726 | Helped |
| Use attentive fusion instead of Perplexity Distill (ablation) | REVEAL-Base retriever training · OK-VQA | accuracy (%): 46.7 -> 47.3 | Helped |
| Use attentive fusion instead of EMDR² (ablation) | REVEAL-Base retriever training · WIT retrieval | Acc@100: 0.869 -> 0.894 | Helped |
| Use attentive fusion instead of ALIGN (fixed) (ablation) | REVEAL-Base retriever training · OK-VQA | accuracy (%): 44.7 -> 47.3 | Helped |
Closest papers by meaning
- Knowledge-based Visual Question Answer with Multimodal Processing, Retrieval and Filtering (NeurIPS 2025 · similarity 0.90)
- MRAG-Bench: Vision-Centric Evaluation for Retrieval-Augmented Multimodal Models (ICLR 2025 · similarity 0.89)
- EVCap: Retrieval-Augmented Image Captioning with External Visual-Name Memory for Open-World Comprehension (CVPR 2024 · similarity 0.89)
- Retrieval-Augmented Visual Question Answering via Built-in Autoregressive Search Engines (AAAI 2025 · similarity 0.89)
- Recoverable Compression: A Multimodal Vision Token Recovery Mechanism Guided by Text Information (AAAI 2025 · similarity 0.89)
- Augmenting Multimodal LLMs with Self-Reflective Tokens for Knowledge-based Visual Question Answering (CVPR 2025 · similarity 0.89)
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.