Super-CLEVR

Super-CLEVR is a virtual benchmark designed to evaluate Visual Question Answering (VQA) models by isolating and studying the effects of various factors of variation in domain shifts, including visual complexity, question redundancy, concept distribution, and concept compositionality.

Papers using Super-CLEVR (3)

Euclid's Gift: Enhancing Spatial Perception and Reasoning in Vision-Language Models via Geometric Surrogate Tasks2025 · 1 citesThink or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Models2025Super-CLEVR: A Virtual Benchmark to Diagnose Domain Robustness in Visual Reasoning2022 · 4 cites