List your company

Long-context language modeling on OpenWebText (val)

3.93NLL (Context=1024)Tensor Cache
3.92083.98294.0454.1071May 21, 2026

Evaluation results

MethodLinks
Tensor CacheParameters=130M, Training iterations=50K, Training block size=1024, Window size (W)=5122026.053.9344.144.294.555.140.8
Full KVParameters=130M, Training iterations=50K, Training block size=1024, Window size (W)=5122026.054.054.464.495.135.461.94
StreamingLLMParameters=130M, Training iterations=50K, Training block size=1024, Window size (W)=5122026.054.114.664.744.855.355.650.76
Infini-attentionParameters=130M, Training iterations=50K, Training block size=1024, Window size (W)=5122026.054.144.054.494.575.015.420.76
Window KVParameters=130M, Training iterations=50K, Training block size=1024, Window size (W)=5122026.054.163.984.864.785.085.440.75