Replication Data for: Q-Mirror: Unlocking the Multi-Modal Potential of Scientific Text-Only QA Pairs

Q-Mirror is a benchmark dataset for evaluating the transformation of text-only scientific QA pairs (TQAs) into high-quality multi-modal QA pairs (MMQAs). It contains 440 expert-verified questions spanning 22 scientific disciplines, partitioned into Q-Mirror-Expert (310 questions) and Q-Mirror-Grad (130 questions). This repository includes JSONL annotation files and generated MMQA images (organized by question ID, PNG format) for the NeurIPS 2026 Evaluations and Datasets Track.