Research map / Vision-language and multimodal / Visual question answering
Maintaining Reasoning Consistency in Compositional Visual Question Answering
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 |
|---|---|---|---|
| Integrate sub-question answering into the compositional-question reasoning process (ablation) | Graph reasoning model on GQA/GQA-Sub · GQA-Sub validation-sub | RC(3): 51.62 -> 52.80 | Helped |
| Add the consistency constraint (ablation) | Dialog-like reasoning model · GQA-Sub validation-sub | RC(3): 51.64 -> 52.80 | Helped |
| Use the ordered sub-question iteration schedule instead of random iteration counts (ablation) | Dialog-like reasoning model · GQA-Sub validation-sub | RC(3): 50.25 -> 52.80 | Helped |
| Add data augmentation using GQA-Sub sub-questions (ablation) | Graph reasoning model · GQA validation | accuracy: 65.29 -> 65.58 | No clear effect |
| Add data augmentation using GQA-Sub sub-questions (ablation) | Graph reasoning model · GQA-Sub validation-sub | RC(3): 33.62 -> 51.62 | Helped |
| Use dialog-like reasoning and the consistency constraint rather than the strongest reported augmented baseline | GQA-Sub validation-sub · GQA-Sub validation-sub | RC(3): 50.05 -> 52.80 | Helped |
| Use the proposed method rather than MDETR | GQA test · GQA test | accuracy: 61.99 -> 59.58 | Hurt |
Closest papers by meaning
- Core-to-Global Reasoning for Compositional Visual Question Answering (AAAI 2025 · similarity 0.93)
- GQA: A New Dataset for Real-World Visual Reasoning and Compositional Question Answering (CVPR 2019 · similarity 0.92)
- Logical Implications for Visual Question Answering Consistency (CVPR 2023 · similarity 0.91)
- ConMe: Rethinking Evaluation of Compositional Reasoning for Modern VLMs (NeurIPS 2024 · similarity 0.89)
- Align and Aggregate: Compositional Reasoning with Video Alignment and Answer Aggregation for Video Question-Answering (CVPR 2024 · similarity 0.88)
- VTQA: Visual Text Question Answering via Entity Alignment and Cross-Media Reasoning (CVPR 2024 · similarity 0.88)
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.