List your company

LLM Performance Estimation on MMLU (test)

0.842MAE (%)SparseEval
0.560082.463044.3666.26896Feb 8, 2026

Evaluation results

MethodLinks
SparseEvalNumber of Anchors=1002026.020.8420.908——————
SparseEvalNumber of Anchors=802026.020.9620.896——————
SparseEvalNumber of Anchors=602026.020.9970.89——————
SparseEvalNumber of Anchors=402026.021.2820.871——————
SparseEvalNumber of Anchors=202026.021.7180.832——————
Anchor PointsNumber of Anchors=802026.021.9640.856——————
TailoredBenchNumber of Anchors=1002026.022.0190.862——————
gp-IRTNumber of Anchors=1002026.022.2020.829——————
TailoredBenchNumber of Anchors=802026.022.2160.876——————
Anchor PointsNumber of Anchors=602026.022.3310.85——————
TailoredBenchNumber of Anchors=602026.022.4210.857——————
gp-IRTNumber of Anchors=802026.022.5370.798——————
TailoredBenchNumber of Anchors=402026.022.6770.845——————
Anchor PointsNumber of Anchors=402026.022.830.801——————
gp-IRTNumber of Anchors=602026.023.190.71——————
gp-IRTNumber of Anchors=402026.023.8020.692——————
TailoredBenchNumber of Anchors=202026.024.0460.755——————
Anchor PointsNumber of Anchors=202026.024.8980.727——————
gp-IRTNumber of Anchors=202026.025.940.569——————
Anchor PointsNumber of Anchors=1002026.027.890.764——————
Domain-Proxy BaselineBase Model=Llama-3-8B-Instruct2026.06——10.010.660.812.17.416.1
Domain-Proxy BaselineBase Model=Qwen2-0.5B2026.06——9.370.730.851016.425.3
Domain-Proxy BaselineBase Model=Qwen2.5-7B-Instruct2026.06——6.580.850.928.62232.8
Early-Dynamics BaselineBase Model=Llama-3-8B-Instruct2026.06——7.330.780.8821.537.748.6
Early-Dynamics BaselineBase Model=Qwen2-0.5B2026.06——7.470.810.919.328.839.7
Early-Dynamics BaselineBase Model=Qwen2.5-7B-Instruct2026.06——3.330.960.9829.95067.5
Early-Stop ExtrapolationBase Model=Llama-3-8B-Instruct2026.06——10.460.660.813.78.817.8
Early-Stop ExtrapolationBase Model=Qwen2-0.5B2026.06——10.410.690.8311.7—18.7
Early-Stop ExtrapolationBase Model=Qwen2.5-7B-Instruct2026.06——7.430.810.911.223.932.8
ProxyLMBase Model=Llama-3-8B-Instruct2026.06——6.760.860.9330.945.862.5
ProxyLMBase Model=Qwen2-0.5B2026.06——40.90.9535.653.772.1
ProxyLMBase Model=Qwen2.5-7B-Instruct2026.06——2.110.980.9940.767.985.8
TUNEAHEAD (Full)Base Model=Llama-3-8B-Instruct2026.06——5.020.860.933655.873.3
TUNEAHEAD (Full)Base Model=Qwen2-0.5B2026.06——3.750.910.9539.258.474.6
TUNEAHEAD (Full)Base Model=Qwen2.5-7B-Instruct2026.06——1.470.990.995082.595.1
TUNEAHEAD-Dynamic-OnlyBase Model=Qwen2.5-7B-Instruct2026.06——3.380.960.9919.835.855.6
TUNEAHEAD-Static-OnlyBase Model=Qwen2.5-7B-Instruct2026.06——3.50.960.9914.932.849.3