Experiments to raise scroot's groundedness correlation from ρ=0.40 toward RAGAS's ρ=0.64, staying deterministic, free, and CPU-runnable. All experiments evaluated on the same 396 SummEval samples used ...
End-to-end quality benchmarks for the scroot library. All benchmarks run locally - no API key required for the core suite (DeepEval/RAGAS comparison optionally requires OPENAI_API_KEY).
Some results have been hidden because they may be inaccessible to you
Show inaccessible results