Research map / Vision-language and multimodal / Visual question answering
ManyModalQA: Modality Disambiguation and QA over Diverse Inputs
Darryl Hannan, Akshay Jain, Mohit Bansal
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 the predicted modality to select a pretrained unimodal QA model | ManyModalQA · ManyModalQA test | accuracy: 21.12% -> 39.73% | Improved |
| Use the ground-truth modality rather than the predicted modality (ablation) | ManyModalQA model · ManyModalQA test | accuracy: 39.73% -> 46.26% | Improved |
| Replace ELMo with RoBERTa in the modality selector (ablation) | Modality selector · ManyModalQA test | accuracy: 77.49% -> 81.35% | Improved |
| Replace ELMo with RoBERTa in the modality selector (ablation) | Modality selector · ManyModalQA dev | accuracy: 77.93% -> 80.92% | Improved |
| Replace ELMo with RoBERTa in the modality selector (ablation) | Modality selector · SQuAD v1.1, VQA v2, and WikiTableQuestions | accuracy: 92.65% -> 95.44% | Improved |
| Replace BERT with RoBERTa in the text QA model (ablation) | Text QA model · ManyModalQA | performance | No clear effect |
| Replace the bottom-up attention image QA model with LXMERT (ablation) | Image QA model · ManyModalQA test, images | accuracy: 27.17% -> 37.11% | Improved |
| Replace the bottom-up attention image QA model with LXMERT (ablation) | Image QA model · ManyModalQA test, images, oracle | accuracy: 29.05% -> 40.79% | Improved |
Closest papers by meaning
- MultiModalQA: complex question answering over text, tables and images (ICLR 2021 · similarity 0.92)
- Assessing Modality Bias in Video Question Answering Benchmarks with Multimodal Large Language Models (AAAI 2025 · similarity 0.90)
- Querying as Prompt: Parameter-Efficient Learning for Multimodal Language Model (CVPR 2024 · similarity 0.88)
- Evaluating Cross-Modal Reasoning Ability and Problem Charactaristics with Multimodal Item Response Theory (ICLR 2026 · similarity 0.87)
- Cross-modal Information Flow in Multimodal Large Language Models (CVPR 2025 · similarity 0.87)
- SK-VQA: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal LLMs (ICML 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.
