Meta-evaluation on AgentEvalBench
83.8URFEvalAgent
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| EvalAgentMeta-Evaluator LLM=Sonnet 4.5, Baseline Comparison=B12026.05 | 83.8 | 90 | 92.5 | — | — | 90 | |
| EvalAgentMeta-Evaluator LLM=Haiku 4.5, Baseline Comparison=B12026.05 | 81.2 | 92.5 | 98.8 | — | — | 95 | |
| EvalAgentMeta-Evaluator LLM=Haiku 4.5, Baseline Comparison=B22026.05 | 80 | 93.8 | 93.8 | — | — | 95 | |
| EvalAgentMeta-Evaluator LLM=Sonnet 4.5, Baseline Comparison=B22026.05 | 76.2 | 85 | 92.5 | — | — | 85 | |
| EvalAgentMeta-Evaluator LLM=Haiku 4.5, Baseline Comparison=B32026.05 | 63.7 | 83.8 | 91.2 | — | — | 85 | |
| EvalAgentMeta-Evaluator LLM=Sonnet 4.5, Baseline Comparison=B32026.05 | 63.7 | 73.8 | 76.2 | — | — | 76.2 | |
| EvalAgentMeta-Evaluator LLM=Sonnet 4.5, Baseline Comparison=B42026.05 | 60 | 81.2 | 87.5 | 73.8 | 56.2 | 85 | |
| EvalAgentMeta-Evaluator LLM=Haiku 4.5, Baseline Comparison=B42026.05 | 51.2 | 85 | 95 | 85 | 53.8 | 90 |