ResearchBenchmarksCrystallize on scientific experiment datasetFollow0.301Success RateCAPER0.10860.158550.20850.25845Feb 10, 2026Evaluation ResultsMethodMethodLinksSuccess RateCAPER2026.020.301MP+MLM+RLBackbone=GPT-4oBackbone=GPT-4o2026.020.215MLM+RLBackbone=GPT-4oBackbone=GPT-4o2026.020.21MLM+RLBackbone=LLaMA3.2, Rea...Backbone=LLaMA3.2, Reasoning Strategy=w/o CoT2026.020.129LLM+RLBackbone=GPT-4oBackbone=GPT-4o2026.020.124LLM+RLBackbone=Llama-3.1-8B,...Backbone=Llama-3.1-8B, Reasoning Strategy=w/o CoT2026.020.116