List your company

Vocabulary Difficulty Prediction on KVL data (test)

0.989PCC (Chinese)Statistical Optimum
0.743560.807280.8710.93472May 14, 2026

Evaluation results

MethodLinks
Statistical OptimumCategory=Upper Bound / Simulation2026.050.9890.990.9950.991
finetuned_llms_plusTrack=Open-track submission, Fine-tuning=Soft-target cross-entropy2026.050.9280.9150.9190.921
open_maxTrack=Open-track submission, Fine-tuning=Soft-target cross-entropy2026.050.9270.9160.920.921
≤32B LLM AverageModel Type=Ensemble2026.050.9260.9140.9160.918
finetuned_llmsTrack=Open-track submission, Fine-tuning=Soft-target cross-entropy2026.050.9250.9140.9150.918
≤14B LLM AverageModel Type=Ensemble2026.050.9210.9050.9080.911
GLM-4-32BModel Size=32B2026.050.9180.9070.9060.91
Qwen2.5-32BModel Size=32B2026.050.9170.9020.9070.909
≤9B LLM AverageModel Type=Ensemble2026.050.9160.9010.90.906
Ministral-3-14BModel Size=14B2026.050.9150.9020.9070.908
closed_maxtrack=closed-track2026.050.8740.8440.8540.857
explainabletrack=closed-track2026.050.8370.7790.7890.802
explainableinference_type=std. inference2026.050.820.7680.7760.788
explainableinference_type=lin. regression2026.050.8150.7660.7690.783
Open-Track BaselineCategory=Shared Task Baseline2026.050.8040.7860.7830.791
traditionaltrack=closed-track2026.050.7670.7470.7210.745
Closed-Track Baselinetrack=closed-track2026.050.7530.7730.7650.764