Ranking Correlation Analysis on ArxivRollBench and ChatbotArena
0.86Spearman Correlation CoefficientA.R.Bench (S) - A.R.Bench (C)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| A.R.Bench (S) - A.R.Bench (C)Comparison Pair=Sequencing vs Cloze2025.07 | 0.86 | 0.92 | 0.77 | |
| A.R.Bench (S) - A.R.Bench (P)Comparison Pair=Sequencing vs Prediction2025.07 | 0.86 | 0.86 | 0.69 | |
| A.R.Bench (C) - A.R.Bench (P)Comparison Pair=Cloze vs Prediction2025.07 | 0.86 | 0.86 | 0.69 | |
| A.R.Bench (S) - ChatbotArenaA.R.Bench Strategy=Sequencing, Reference Benchmark=ChatbotArena2025.07 | 0.76 | 0.71 | 0.6 | |
| A.R.Bench (P) - ChatbotArenaA.R.Bench Strategy=Prediction, Reference Benchmark=ChatbotArena2025.07 | 0.73 | 0.69 | 0.55 | |
| A.R.Bench (C) - ChatbotArenaA.R.Bench Strategy=Cloze, Reference Benchmark=ChatbotArena2025.07 | 0.61 | 0.51 | 0.55 |