Model Arena · Evaluation history
Every pass you've run, remembered.
CineQC records each finished evaluation here — single passes and batch triage alike. Group by model, watch scores trend over time, and export the history as data.
Stored in this browser only — nothing uploads, ever. Clearing site data clears the Arena.Run a clip (or a batch) through CineQC and it lands here automatically. Tip: name files model_clipname.mp4 — e.g. kling-16_stadium.mp4 — and the Arena groups them by model on its own. Click any model name later to rename it.
Inter-annotator agreement
Score the same clip in CineQC as two different evaluators, copy each pass's shareable link, and paste both here. You get the per-category deltas and a chance-corrected reliability score — the inter-rater consistency human-feedback pipelines are built on.
This compares one clip. To trend agreement across a whole set and see which category the passes split on, use the rubric reliability study.
Leaderboard
Avg is the mean automated score (0–100 across the five metrics). Rubric avg and pass rate only count runs you scored by hand. Click a model name to rename it.
Run log