LMArena
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
LMArena-based simulation
-2.5Top-8 Ranked Candidates
9
LMArena (held-out models)
0.6Beta Coefficient
8
LMArena 140K
0.213Kappa (κ)
7
LMArena
50Win Rate vs. Generative UI
4
LMArena Expert Med
0.476Kendall's tau_b
3
LMArena Med
0.594Kendall's Tau_b
3