\ud83e\uddea ChemVQA-2K: A Visual Question Answering Dataset for Molecular Understanding

\ud83e\uddea ChemVQA-2K: A Visual Question Answering Dataset for Molecular Understanding\ud83d\udcd8 OverviewChemVQA-2K is a novel Visual Question Answering (VQA) dataset designed to bridge chemistry and multimodal AI.It contains approximately 2,000 high-resolution molecular images (512\u00d7512) generated from valid SMILES strings, accompanied by 10 structured Q&A pairs per molecule, resulting in ~20,000 image-question-answer triplets.Each image represents a 2D chemical structure rendered using RDKit, while each question tests the model\u2019s ability to reason over molecular features such as formula, atom counts, bonds, functional groups, and polarity. \ud83e\uddec Dataset StructureComponentDescriptionChemVQA_2K_images.zip 2,000 molecule renderings (mol_0.png, mol_1.png, \u2026)ChemVQA_2K_full.csvComplete dataset with columns: id, image_name, question, answer  Each record follows:{  \id\: \mol_123\,  \image_name\: \mol_123.png\, \question\: \What is the molecular formula of this molecule?\,  \answer\: \C6H6O2\} \ud83d\udd0d Example QuestionsEach molecule has multiple Q&A pairs, e.g.:QuestionExample AnswerWhat is the molecular formula of this molecule?C\u2082H\u2085OHWhat is the molecular weight?46.07 g\/molHow many total atoms are present?9Which functional groups are present?AlcoholIs the molecule polar or non-polar?Polar \u2699\ufe0f Data Generation ProcessMolecules generated by concatenating random organic fragments and validated using RDKit.Each molecule\u2019s image created with Draw.MolToFile() at 512\u00d7512 px resolution.Functional groups detected via SMARTS pattern matching.Q&A pairs auto-generated from chemical descriptors (MolWt, CalcMolFormula, substructure matches). \ud83d\ude80 Intended UseChemVQA-2K is ideal for:Fine-tuning Vision-Language Models (VLMs) for scientific visual reasoning.Developing chemistry-aware question answering systems.Training vision encoders on molecular visual patterns.Exploring RL-based visual understanding of chemical structures. \ud83d\udcca Dataset StatisticsPropertyValueImages1924Image resolution512\u00d7512 pxQ&A pairs19240Functional groups detected16File size (approx.)~25 MB (images + CSVs) \ud83e\udde0 Potential Research DirectionsMultimodal Chemistry Understanding \u2014 connecting visual structure with symbolic reasoning.Scientific Vision-Language Pretraining \u2014 use as domain-specific VQA benchmark.Explainable Chemistry AI \u2014 models that describe functional features and molecular properties.