LLM-as-a-Judge
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Multi-fidelity bandit optimization | LLM-as-a-judge residual-mismatch Λ=128000 (test) | Mean Cost-Weighted Pseudo-Regret4,023.4 | 4 | |
| Preference evaluation | LLM-as-a-Judge comparison set | TCRM Better Rate33.7 | 1 |
Showing 2 of 2 rows