NRS Navigator · blinded relevance review

Judge the sources before seeing scores

Randomized source order. No system ranks, model scores or proposed labels are shown. Grade the full section separately from the passage, caption and heading supplied to Jev. Grades: 0 unrelated; 1 topical but not useful; 2 useful support; 3 directly addresses this query. Leave uncertain items unreviewed.

Download labels to retain your work; this page also keeps a local browser draft when storage is available. Export only includes explicitly completed reviews. Import the downloaded JSON with scripts/score_benchmark.py --labels PATH to calculate independent metrics without additional Jev calls.