Research map / Vision-language and multimodal / Visual question answering
What Value Do Explicit High Level Concepts Have in Vision to Language Problems?
Check 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 |
|---|---|---|---|
| Add a predicted semantic-attribute layer between image features and the LSTM | CNN-RNN image captioning · MS COCO test | BLEU-1: 0.68 -> 0.74 | Helped |
| Add a predicted semantic-attribute layer between image features and the LSTM | CNN-RNN image captioning · MS COCO test | CIDEr: 0.73 -> 0.94 | Helped |
| Use predicted semantic attributes instead of fine-tuned CNN features | Visual question answering · Toronto COCO-QA | accuracy: 58.34 -> 61.38 | Helped |
| Use predicted semantic attributes instead of CNN image features | Visual question answering · VQA test-standard | accuracy: 54.06 -> 55.84 | Helped |
| Expand image-sourced attributes with WordNet knowledge | Att-CNN+LSTM · VQA test-standard | accuracy: 55.84 -> 57.62 | Helped |
| Expand image-sourced attributes with WordNet knowledge | Att-CNN+LSTM · VQA validation, why questions | accuracy: 7.77 -> 9.88 | Helped |
| Replace predicted attributes with ground-truth attributes (ablation) | Att-CNN+LSTM · MS COCO test | BLEU-1: 0.74 -> 0.80 | Helped |
| Replace CNN-predicted attributes with SVM-predicted attributes (ablation) | Att-CNN+LSTM · MS COCO test | BLEU-1: 0.74 -> 0.69 | Hurt |
| Combine CNN features with the attribute vector instead of using attributes alone (ablation) | Att-CNN+LSTM · MS COCO test | BLEU-1: 0.74 -> 0.71 | Hurt |
| Use predicted attributes instead of ground-truth attributes (ablation) | Att-GT+LSTM · VQA July Release, overall | accuracy: 54.10 -> 51.60 | Hurt |
Closest papers by meaning
- Aligning Visual Regions and Textual Concepts for Semantic-Grounded Image Representations (NeurIPS 2019 · similarity 0.89)
- Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.89)
- Language-Informed Visual Concept Learning (ICLR 2024 · similarity 0.89)
- Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models (NeurIPS 2025 · similarity 0.88)
- Open-ended VQA benchmarking of Vision-Language models by exploiting Classification datasets and their semantic hierarchy (ICLR 2024 · similarity 0.88)
- From Head to Tail: Towards Balanced Representation in Large Vision-Language Models through Adaptive Data Calibration (CVPR 2025 · 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.