AINeutralarXiv – CS AI · Mar 116/10
🧠
OPENXRD: A Comprehensive Benchmark Framework for LLM/MLLM XRD Question Answering
Researchers introduced OPENXRD, a comprehensive benchmarking framework for evaluating large language models and multimodal LLMs in crystallography question answering. The study tested 74 state-of-the-art models and found that mid-sized models (7B-70B parameters) benefit most from contextual materials, while very large models often show saturation or interference.
🧠 GPT-4🧠 GPT-4.5🧠 GPT-5