Task gallery

100 tasks across 4 families.

Each task: a realistic brief, a source video, a verifiable rubric. Below: a sample. Full task set on Hugging Face.

Assembly

18

tasks

Repair

18

tasks

Sequencing

28

tasks

Repurpose

36

tasks

Assembly

ASM-007

5-slot narrative storyboard, AI film

Source length
3:42
Distractors per slot
2
Best agent
GPT-5.5 · Codex
Best score
42%

Repair

RPR-012

Color grade drift across 3 cuts

Source length
1:18
Best agent
GPT-5.5 · OpenCode
Best score
51%

Sequencing

SEQ-019

12-clip cinematic short, shuffled

Source length
5:04
Best agent
GPT-5.5 · Codex
Best score
38%

Repurpose

RPS-024

11-min political talk → 60s repurposed cut

Source length
11:00
Best agent
GPT-5.5 · OpenClaw
Best score
26%

Assembly

ASM-014

4-slot commercial storyboard

Source length
2:08
Distractors per slot
2
Best agent
Gemini 3.1 Pro · Gemini CLI
Best score
50%

Repurpose

RPS-031

Music performance → 45s highlight

Source length
8:15
Best agent
GPT-5.5 · OpenCode
Best score
22%