File size: 3,146 Bytes
f517c0e
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
{
 "suite": "robopaint",
 "task": "oil_starrynight_i01",
 "source": "oracle",
 "kind": "teaser",
 "renderer": "SAPIEN 3.0.3 'rt' ray-tracing shader pack (ManiSkill 3.0.1), spp=128, path_depth=8, denoiser=optix (OptiX), rendered natively at 2560x1440; one soft area light + ambient, dark board",
 "rendering": "ray-tracing",
 "width": 2560,
 "height": 1440,
 "fps": 30,
 "frames": 765,
 "frames_painting": 660,
 "duration_s": 25.5,
 "speed": 119.21,
 "speed_note": "average simulated seconds per video second; the time-lapse is path-uniform (equal inked-path length per frame), not constant-speed",
 "time_lapse": "path-uniform: one frame per 1/660 of the inked path (20205 mm), none during travel or dips; 0.5 s hold on the bare sheet, a short arm-retreat tail, 2.5 s hold on the finished painting",
 "sim_seconds": 3039.8,
 "control_steps": 60797,
 "strokes": 1682,
 "seconds_per_frame": 0.6269,
 "render_wall_s": 488.6,
 "success_check": true,
 "success_check_how": "the reference solution (solution/oracle_qpos.npy, as solution/oracle_paint.py replays it) was written to a trajectory.npz and graded by the task's own verifier, tests/grade.py --repeat 2 (two fresh replay processes with tests/rcb_brush_strict.py on the graders' frozen tests/instance.json; only its optional replay video was switched off): success=1 deterministic=1 instance_ok=1 final_reward=0.70468. The un-rendered replay with the same tooling scored success=1.0; the rendered replay (same tooling and instance, physics untouched: only a camera, lights and materials added) scored success=1.0 on its own final state, and its state hash c0dc305b6a9d0ba9 and main-sheet canvas equal the un-rendered replay's (True, True)",
 "checks": {
  "verifier_success": true,
  "verifier_deterministic": true,
  "verifier_instance_ok": true,
  "passA_success": true,
  "passA_instance_exact": true,
  "render_success": true,
  "render_state_sha_equals_passA": true,
  "render_canvas_equals_passA": true,
  "render_instance_exact": true
 },
 "verifier_reward": {
  "success": 1,
  "final_reward": 0.70468,
  "deterministic": 1,
  "instance_ok": 1,
  "instance_exact": 1,
  "strict_ok": 1.0,
  "image_success": 1.0,
  "path_mm": 20746.89,
  "n_strokes": 1682.0,
  "n_actions": 60797,
  "de_mean": 5.9064,
  "ssim_L": 0.720901,
  "paint_iou": 0.999741,
  "iou": null
 },
 "state_sha": "c0dc305b6a9d0ba9",
 "task_dir": "the exam's task repository@d43e2ae4b:tasks/robopaint-oil-starrynight-i01-privileged",
 "image": "eai-robopaint:0.2.0 (manifest sha256:f7f925094d394989362051da1605edb4f23645732064f93e9726d704c89deb12, config sha256:0d047bf5fdc70613396b4074c1339d39c44c522c46f44a7b5f5b0bc80cfd9353)",
 "created": "2026-10-16T01:40:47-05:00",
 "encode": "libx264 High yuv420p CRF 16 preset slow, 30 fps constant, bt709, faststart",
 "files": {
  "teaser.mp4": {
   "codec": "h264",
   "profile": "High",
   "pix_fmt": "yuv420p",
   "width": 2560,
   "height": 1440,
   "r_frame_rate": "30/1",
   "frames": 765,
   "duration": 25.5,
   "size": 19474646
  }
 },
 "camera": "keyframe path start -> ground -> forms -> pull-back -> end (whole arm), cubic Hermite easing over the painting frames"
}