Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Held-out Elo Estimation on LMArena (held-out models)
Loading...
0.6
Beta Coefficient
Soft-Elo
0.3504
0.4152
0.48
0.5448
Jun 11, 2026
Beta Coefficient
Hard Elo MAE
Soft Elo MAE
Delta MAE (%)
Delta Correlation (Δρ)
Updated 1mo ago
Evaluation Results
Method
Method
Links
Beta Coefficient
Hard Elo MAE
Soft Elo MAE
Delta MAE (%)
Delta Correlation (Δρ)
Soft-Elo
Judge=Qwen3.5-27B
2026.06
0.6
46
13.6
70
0.003
Soft-Elo
Judge=GPT-OSS-120B
2026.06
0.58
47.4
14.4
70
0.002
Soft-Elo
Judge=Gemma4-26B-A4B
2026.06
0.54
55.9
15.7
72
0.011
Soft-Elo
Judge=Qwen3-32B
2026.06
0.5
27.5
16.7
39
0.005
Soft-Elo
Judge=GPT-OSS-20B
2026.06
0.46
34.5
19.8
42
0.009
Soft-Elo
Judge=Llama-3.3-70B
2026.06
0.41
43.9
24.5
44
0.004
Soft-Elo
Judge=Gemma4-E4B
2026.06
0.38
48.2
21
57
0.012
Soft-Elo
Judge=DeepSeek-V3.2
2026.06
0.36
63.4
17.1
73
0.003
Feedback
Search any
task
Search any
task