Research map / Vision-language and multimodal / Multimodal LLMs
V-PROM: A Benchmark for Visual Reasoning Using Visual Progressive Matrices
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 pairwise relational reasoning instead of a non-relational MLP | V-PROM models with ResNet features · V-PROM neutral | accuracy: 41.2 -> 51.2 | Helped |
| Use pairwise relational reasoning instead of a non-relational MLP | V-PROM models with Bottom-Up features and auxiliary loss · V-PROM neutral | accuracy: 47.7 -> 61.3 | Helped |
| Add relationship-type auxiliary supervision (ablation) | RN with ResNet features · V-PROM neutral | accuracy: 51.2 -> 55.8 | Helped |
| Add relationship-type auxiliary supervision (ablation) | RN with Bottom-Up features · V-PROM neutral | accuracy: 55.4 -> 61.3 | Helped |
| Add relationship-type auxiliary supervision (ablation) | RN, ResNet · V-PROM interpolation | accuracy: 39.8 -> 42.4 | Helped |
| Add relationship-type auxiliary supervision (ablation) | RN, ResNet · V-PROM extrapolation | accuracy: 39.0 -> 42.3 | Helped |
| Add relationship-type auxiliary supervision (ablation) | RN, Bottom-Up · V-PROM held-out relationship/attribute pairs | accuracy: 42.2 -> 44.9 | Helped |
| Use Bottom-Up Attention features instead of ResNet features (ablation) | RN · V-PROM neutral | accuracy: 51.2 -> 55.4 | Helped |
| Use Bottom-Up Attention features instead of ResNet features (ablation) | RN with relationship-type auxiliary loss · V-PROM neutral | accuracy: 55.8 -> 61.3 | Helped |
| Remove panel-ID representations (ablation) | RN with ResNet features · V-PROM neutral | accuracy: 32.6 vs 51.2 | Hurt |
| Remove panel-ID representations (ablation) | RN with Bottom-Up features and auxiliary loss · V-PROM neutral | accuracy: 38.0 vs 61.3 | Hurt |
| Shuffle context panels (ablation) | RN · V-PROM neutral | accuracy: 51.2 -> 12.5 | Hurt |
Closest papers by meaning
- VisRes Bench: On Evaluating the Visual Reasoning Capabilities of VLMs (CVPR 2026 · similarity 0.92)
- Benchmarking Multimodal Mathematical Reasoning with Explicit Visual Dependency (NeurIPS 2026 · similarity 0.91)
- Not Another Text Benchmark: Putting the “Visual" Back in Visual Question Answering for Large Video Models (NeurIPS 2026 · similarity 0.90)
- VGR: Visual Grounded Reasoning (ICLR 2026 · similarity 0.90)
- Multi-step Visual Reasoning with Visual Tokens Scaling and Verification (NeurIPS 2025 · similarity 0.90)
- Visual Graph Arena: Evaluating Visual Conceptualization of Vision and Multimodal Large Language Models (ICML 2025 · similarity 0.90)
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.