ResearchBenchmarksLong-Context Language Modeling on RULER (Context Sweep)Long-Context Language Modeling on RULER (Context Sweep)Follow81.5Accuracy (8k Context)CoPE80.480880.745481.0181.2746Feb 5, 2026Evaluation resultsMethodMethodLinksAccuracy (8k Context)Accuracy (16k Context)Accuracy (32k Context)Accuracy (64k Context)Accuracy (128k Context)Accuracy (256k Context)CoPE2026.0281.582.8482.7576.7161.9546.86RoPE2026.0280.5282.3382.1176.9361.1928.86