C-VQA

C-VQA is a novel dataset designed to evaluate the counterfactual reasoning capabilities of multi-modal large language models by incorporating original questions with counterfactual presuppositions across various types and difficulty levels.

Papers using C-VQA (8)

Finding Culture-Sensitive Neurons in Vision-Language Models2025FRAMES-VQA: Benchmarking Fine-Tuning Robustness across Multi-Modal Shifts in Visual Question Answering2025Greedy Gradient Ensemble for Robust Visual Question Answering2021 · 5 citesSC-ML: Self-supervised Counterfactual Metric Learning for Debiased Visual Question Answering2023 · 2 citesOn the Flip Side: Identifying Counterexamples in Visual Question Answering2018Taking a HINT: Leveraging Explanations to Make Vision and Language Models More Grounded2019What If the TV Was Off? Examining Counterfactual Reasoning Abilities of Multi-modal Language Models2023WeaQA: Weak Supervision via Captions for Visual Question Answering2020