#crystallography News & Analysis

2 articles tagged with #crystallography. AI-curated summaries with sentiment analysis and key takeaways from 50+ sources.

2 articles

AINeutralarXiv – CS AI · May 296/10

🧠

CrystalXRD-Bench: Benchmarking Vision-Language Models for XRD Peak Indexing Across Diverse Crystalline Materials

Researchers introduced CrystalXRD-Bench, a 250-sample benchmark dataset for evaluating vision-language models on crystallographic peak indexing from X-ray diffraction patterns. Despite testing seven leading VLMs, the best model achieved only 37.6% exact-match accuracy, revealing significant gaps in how AI systems handle precise scientific figure interpretation and multi-step reasoning.

🧠 GPT-5

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