Dataset: datasets/rag_qa_small.jsonl
| Candidate | Quality | Grounded | Citation | Latency ms | Cost/query | Score | Constraints |
|---|---|---|---|---|---|---|---|
| baseline-lite | 0.333 | 0.907 | 1.000 | 260.2 | 0.00004 | 0.498 | PASS |
| candidate-detailed | 0.281 | 0.684 | 1.000 | 178.7 | 0.00011 | 0.427 | PASS |
Winner: baseline-lite
baseline-lite has the highest score among candidates that passed hard constraints.
Baseline: baseline-lite vs Candidate: candidate-detailed
| Metric | Delta |
|---|---|
| quality | -0.0526 |
| groundedness | -0.2231 |
| citation | +0.0000 |
| latency_ms | -81.57 |
| cost | +0.000074 |
| score | -0.0708 |
Verdict: FAIL