Research map / Vision-language and multimodal / Visual question answering
Transparency by Design: Closing the Gap Between Performance and Interpretability in Visual Reasoning
David Mascharka, Philip Tran, Ryan Soklaski, Arjun Majumdar
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 |
|---|---|---|---|
| Add attention-mask regularization (ablation) | TbD-net · CLEVR | overall accuracy: 98.7 -> 98.5 | Hurt |
| Add attention-mask regularization (ablation) | TbD-net · CLEVR attention interpretability | recall: 0.86 -> 0.92 | Improved |
| Add attention-mask regularization (ablation) | TbD-net · CLEVR attention interpretability | precision: 0.41 -> 0.90 | Improved |
| Use 28×28 rather than 14×14 feature maps (ablation) | TbD + reg · CLEVR | overall accuracy: 98.5 -> 99.1 | Improved |
| Use 28×28 rather than 14×14 feature maps (ablation) | TbD + reg · CLEVR attention interpretability | recall: 0.92 -> 0.99 | Improved |
| Use 28×28 rather than 14×14 feature maps (ablation) | TbD + reg · CLEVR attention interpretability | precision: 0.90 -> 0.98 | Improved |
| Train only on Condition A (ablation) | TbD + reg · CLEVR-CoGenT Condition B | accuracy: 98.8 -> 75.4 | Hurt |
| Fine-tune on 3k Condition B images and 30k questions (ablation) | TbD + reg · CLEVR-CoGenT Condition B | accuracy: 75.4 -> 96.3 | Improved |
| Fine-tune on 3k Condition B images and 30k questions (ablation) | TbD + reg · CLEVR-CoGenT Condition A | accuracy: 98.8 -> 96.9 | Mixed |
| Use TbD + reg + high-resolution feature maps instead of TbD-net (ablation) | TbD-net · CLEVR | overall accuracy: 98.7 -> 99.1 | Improved |
Closest papers by meaning
- Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.90)
- VisRes Bench: On Evaluating the Visual Reasoning Capabilities of VLMs (CVPR 2026 · similarity 0.89)
- Learning to Reason Iteratively and Parallelly for Complex Visual Reasoning Scenarios (NeurIPS 2024 · similarity 0.89)
- No Labels, No Problem: Training Visual Reasoners with Multimodal Verifiers (ICLR 2026 · similarity 0.89)
- CodeDance: A Dynamic Tool-integrated MLLM for Executable Visual Reasoning (CVPR 2026 · similarity 0.89)
- VGR: Visual Grounded Reasoning (ICLR 2026 · 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.
