Showing cs.CVShow all
2 papers · 1 filter
cs.CV2026
When Vision Overrides Language: Evaluating and Mitigating Counterfactual Failures in VLAs
Yu Fang, Yuchun Feng, Dong Jing +5
Vision-Language-Action models (VLAs) promise to ground language instructions in robot control, yet in practice often fail to faithfully follow language. When presented with instruc…
cs.CV2025
GeoBench: Rethinking Multimodal Geometric Problem-Solving via Hierarchical Evaluation
Yuan Feng, Yue Yang, Xiaohan He +8
Geometric problem solving constitutes a critical branch of mathematical reasoning, requiring precise analysis of shapes and spatial relationships. Current evaluations of geometric…