VQA

The VQA dataset is used to evaluate visual question answering tasks, where models are tested on their ability to answer questions about images.

Papers using VQA (3)

Attend, Transform, or Silence: Operator-Level Visual Skipping for Efficient Multimodal LLM Inference2026Prolonged Reasoning Is Not All You Need: Certainty-Based Adaptive Routing for Efficient LLM/MLLM Reasoning2025An Empirical Study on Prompt Compression for Large Language Models2025