List your company

Language Modeling on WikiText (PPL, Latency, and KL Divergence)

0KL DivergenceOracle
-0.2475481.4234013.094354.765299Jun 5, 2026

Evaluation results

MethodLinks
OracleTransfer Pair=MistralLite→Mistral-7B, Network Bandwidth (Bnet)=200 Gbps2026.0604.150331.91
SCDTransfer Pair=MistralLite→Mistral-7B, Network Bandwidth (Bnet)=200 Gbps2026.060.11054.6587.31125.22.65
DroidSpeak-styleTransfer Pair=MistralLite→Mistral-7B, Network Bandwidth (Bnet)=200 Gbps2026.060.14595.33127.34198.71.67
SCD (Reuse-only)Transfer Pair=MistralLite→Mistral-7B, Network Bandwidth (Bnet)=200 Gbps2026.060.16616.1169.02118.22.81
Raw KV (BF16)Transfer Pair=MistralLite→Mistral-7B, Network Bandwidth (Bnet)=200 Gbps2026.063.1209100115.02145.52.28
Quantized (4-bit)Transfer Pair=MistralLite→Mistral-7B, Network Bandwidth (Bnet)=200 Gbps2026.066.188710028.7659.25.6