ResearchBenchmarksLong Context on NIAHFollow99.8AccuracyLlama3.174.00880.70487.494.096Dec 31, 2025Evaluation ResultsMethodMethodLinksAccuracyLlama3.1Size=8B, Type=BaseSize=8B, Type=Base2025.1299.8Gemma3Size=4B, Type=BaseSize=4B, Type=Base2025.1299.5Youtu-LLMSize=2B, Type=BaseSize=2B, Type=Base2025.1298.8Qwen3Size=4B, Type=BaseSize=4B, Type=Base2025.1283Qwen3Size=1.7B, Type=BaseSize=1.7B, Type=Base2025.1279.8SmolLM3Size=3B, Type=BaseSize=3B, Type=Base2025.1275