ResearchBenchmarksMultiple-choice question answering on TIMESERIESEXAMFollow78AccuracyMhybrid38.58448.81759.0569.283Jun 15, 2026Evaluation ResultsMethodMethodLinksAccuracyMhybridModel=gpt-oss-120bModel=gpt-oss-120b2026.0678McodeModel=gpt-oss-120bModel=gpt-oss-120b2026.0670.4MhybridModel=qwen3-next-80bModel=qwen3-next-80b2026.0668.1MdirectModel=gpt-oss-120bModel=gpt-oss-120b2026.0665.3MdirectModel=qwen3-next-80bModel=qwen3-next-80b2026.0663.5McodeModel=qwen3-next-80bModel=qwen3-next-80b2026.0656.4random2026.0640.1