Jev RAG Benchmark

Free English benchmark of an open System One model as a reranking and decision layer in RAG.

Reranking (frozen top-20 candidates)

DatasetnMethodnDCG@10Recall@5MRR@10Rerank p50
scifact300A — no reranker (hybrid order)71.67%79.67%68.75%0 ms
scifact300T — TypeSafe Jev 1.13 batch noul79.29%85.67%77.03%4044 ms
scifact300N — NVIDIA cross-encoder reranker78.70%87.33%76.16%307 ms
xquad-en1190A — no reranker (hybrid order)98.11%99.58%97.60%0 ms
xquad-en1190T — TypeSafe Jev 1.13 batch noul98.93%99.66%98.67%4000 ms
xquad-en1190N — NVIDIA cross-encoder reranker99.37%99.66%99.27%409 ms

Probability calibration (candidate-level relevance)

DatasetModelECE (10 bin)BrierTop-1 accuracyTop-1 confidence (correct)Top-1 confidence (wrong)
scifactTypeSafe Jev 1.13 batch noul0.06250.036671.00%0.8120.549
xquad-enTypeSafe Jev 1.13 batch noul0.01330.004597.90%0.9600.740

Frozen-context answer generation

DatasetGeneratorToken F1Exact matchF1 >= 0.5AbstentionValid citations
xquad-enTdiffusiongemma-26b30.58%1.26%17.31%2.61%99.85%

RAG optimization mode (confidence-partitioned Jev, fixed t = 0.50)

DatasetModelThresholdA baseline nDCG@10Always-on nDCG@10Partitioned nDCG@10Delta vs baseline95% CI
scifactTypeSafe Jev 1.13 batch noul0.5071.67%79.29%75.81%+4.14 pts+2.13 to +6.29
xquad-enTypeSafe Jev 1.13 batch noul0.5098.11%98.93%99.00%+0.89 pts+0.45 to +1.36