Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Long-context Retrieval on RULER niah_single_2 zeroshot (test)

95.5Accuracy

PhiMoE (FP16 baseline)

94.9895.11595.2595.385Apr 20, 2026
Updated 1mo ago

Evaluation Results

MethodLinks
2026.04
95.5---
2026.04
9521.7737.49-0.52