ERQA

What problem does it solve?

This Skill provides a reusable benchmark-dataset card for ERQA, enabling models to evaluate spatial reasoning, world-knowledge understanding, and embodied visual grounding.

Core Features & Use Cases

• Multimodal Question-Answering: Evaluates models on answering questions based on both text and images.
• Spatial Reasoning: Tests models' ability to reason about spatial relations and positions.
• World Knowledge: Assesses models on combining real-world semantics with visual evidence.
• Embodied Reasoning: Evaluates models on movement, viewpoint changes, and visibility.
• Use Case: For a model to assess its capabilities in multi-image reasoning and spatial relations, it can use the ERQA dataset as a benchmark.

Quick Start

Run the ERQA Skill to evaluate your model's performance on spatial reasoning and multimodal question-answering tasks.