Judge alignment prediction on Arena Hard English v2.0
55.78Coding AccuracyTRACE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TRACE2026.05 | 55.78 | 50 | 64.37 | |
| MTLD2026.05 | 52.96 | 53.45 | 58.62 | |
| Perplexity2026.05 | 49.21 | 47.79 | 50.57 | |
| Token Length2026.05 | 44.22 | 47.85 | 36.78 |
| Method | Links | |||
|---|---|---|---|---|
| TRACE2026.05 | 55.78 | 50 | 64.37 | |
| MTLD2026.05 | 52.96 | 53.45 | 58.62 | |
| Perplexity2026.05 | 49.21 | 47.79 | 50.57 | |
| Token Length2026.05 | 44.22 | 47.85 | 36.78 |