Relevance Correlation Evaluation on Logs (All)
0.438Spearman RhoBehavior-Grounded LLM judge
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Behavior-Grounded LLM judgeJudge Type=Behavior-Grounded2026.07 | 0.438 | 0.354 | |
| Plain LLM judgeJudge Type=Plain2026.07 | 0.416 | 0.336 |
Loading the SOTA2 catalog…