Session-level Correlation on MT-Bench
0.74Spearman Correlation (ρ)SKG-Eval
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| SKG-Eval2026.05 | 0.74 | — | 0.56 | |
| SKG-Eval2026.05 | 0.73 | 0.76 | — | |
| GPT-4o-Judgemode=history-aware2026.05 | 0.69 | — | 0.51 | |
| GPT-4o-JudgeType=history-aware2026.05 | 0.66 | 0.69 | — | |
| GPT-4o-Judgemode=turn-only2026.05 | 0.63 | — | 0.47 | |
| DeepEvalmodel=GPT-4o2026.05 | 0.61 | — | 0.45 | |
| DeepEval + GPT-4oEvaluator=DeepEval + GPT-4o2026.05 | 0.59 | 0.62 | — | |
| GPT-4o-JudgeType=turn-only2026.05 | 0.57 | 0.6 | — | |
| LLM-Eval2026.05 | 0.54 | — | 0.39 | |
| LLM-Eval2026.05 | 0.53 | 0.56 | — | |
| ECoh2026.05 | 0.51 | — | 0.37 | |
| ECoh2026.05 | 0.49 | 0.52 | — |