List your company

Inference Efficiency on TQA

0.05Relative Execution Time OverheadPerplexity
-0.238721.710143.6595.60786Mar 20, 2026

Evaluation results

MethodLinks
PerplexityModel=Qwen2.5-14B2026.030.05
PerplexityModel=Llama-2-7B2026.030.066
PerplexityModel=Qwen2.5-7B2026.030.094
PerplexityModel=Qwen2.5-3B2026.030.096
PerplexityModel=Mistral-7B2026.030.102
PerplexityModel=Llama-3-8B2026.030.13
STCModel=Llama-2-7B2026.032.329
STCModel=Mistral-7B2026.032.908
STCModel=Qwen2.5-14B2026.035.866
STCModel=Llama-3-8B2026.036.833
STCModel=Qwen2.5-3B2026.036.946
STCModel=Qwen2.5-7B2026.037.268