Human Alignment Correlation on LMSYS Search Arena (test)
0.74Pearson CorrelationDR-Arena
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DR-ArenaN=62026.01 | 0.74 | 0.94 | |
| BrowseCompN=52026.01 | 0.7 | 0.76 | |
| DeepResearch Bench (Du et al.)N=42026.01 | 0.63 | 0.4 | |
| LiveNewsBenchN=42026.01 | 0.46 | 0.2 | |
| LiveSearchBenchN=42026.01 | -0.23 | -0.11 | |
| LiveResearchBenchN=42026.01 | -0.47 | -0.63 | |
| Deep Research Bench (FutureSearch et al.)N=62026.01 | -0.86 | -0.9 |