A reference-free LLM-judge framework measures the consistency, complexity, and policy coverage of conversational-agent benchmarks, validated with perturbations and human ratings.
Biometrika , volume =
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Benchmarking the Benchmarks: Evaluating Benchmarks for Conversational Agents
A reference-free LLM-judge framework measures the consistency, complexity, and policy coverage of conversational-agent benchmarks, validated with perturbations and human ratings.