List your company

Language Modeling on Shakespeare 1M params (val)

4.79NLLTensor Cache
4.73085.13045.535.9296May 21, 2026

Evaluation results

MethodLinks
Tensor Cachecontext length L=2562026.054.79
Infini-attentioncontext length L=2562026.054.8
Full KVcontext length L=2562026.054.91
Window KVcontext length L=2562026.054.91
StreamingLLMcontext length L=2562026.054.91
Tensor Cachecontext length L=10242026.054.96
Infini-attentioncontext length L=10242026.055.14
Full KVcontext length L=10242026.055.24
Tensor Cachecontext length L=163842026.055.32
StreamingLLMcontext length L=10242026.055.47
Window KVcontext length L=10242026.055.48
Infini-attentioncontext length L=163842026.055.51
Tensor Cachecontext length L=20482026.055.6
Tensor Cachecontext length L=40962026.055.66
Tensor Cachecontext length L=81922026.055.7
Full KVcontext length L=163842026.055.73
StreamingLLMcontext length L=163842026.055.79
Infini-attentioncontext length L=20482026.055.79
Window KVcontext length L=163842026.055.82
Infini-attentioncontext length L=40962026.055.87
Infini-attentioncontext length L=81922026.055.92
Tensor Cachecontext length L=5122026.055.98
Full KVcontext length L=20482026.056.02
Infini-attentioncontext length L=5122026.056.06
Full KVcontext length L=40962026.056.09
Full KVcontext length L=5122026.056.1
Full KVcontext length L=81922026.056.13
Window KVcontext length L=20482026.056.16
StreamingLLMcontext length L=20482026.056.16
StreamingLLMcontext length L=81922026.056.21
Window KVcontext length L=81922026.056.24
StreamingLLMcontext length L=40962026.056.24
Window KVcontext length L=5122026.056.25
StreamingLLMcontext length L=5122026.056.25
Window KVcontext length L=40962026.056.27