List your company

CoT Length Prediction on GPQA Diamond

27.32Relative MAEFuel Gauge
26.02634.760543.49552.2295Mar 11, 2026

Evaluation results

MethodLinks
Fuel GaugeBackbone Model=Qwen3-8B2026.0327.32
Fuel GaugeBackbone Model=Qwen3-4B2026.0331.85
DirectBackbone Model=Qwen3-4B2026.0349.32
MeanBackbone Model=Qwen3-8B2026.0350.03
MedianBackbone Model=Qwen3-8B2026.0352.12
MeanBackbone Model=Qwen3-4B2026.0357.21
DirectBackbone Model=Qwen3-8B2026.0357.95
MedianBackbone Model=Qwen3-4B2026.0359.67