Artifacts
What each model built on each task, shown on the held-out cases it was graded on.
LVGL redraws
Each model rewrote LVGL's software renderer to draw identical frames in fewer RISC-V instructions; the original and each model's best and median renderers are re-run here, showing which pixels each one writes and what each frame costs. Hidden test scenes are the ones the reward is computed on; public dev scenes are ones agents could test on, so they are not graded.