Real Unreal Bench · Task 14
Drop an event and act on playtest notes in an athletics game
Qwen3.8 MaxRun 1Fail
- Outcome
- Fail
- Duration
- 1h 28m
- Steps
- 123
- Tokens
- 24.5M
- Cost
- $8.02
Task
An athletics game drops one of its five events: the agent removes it for good, from the menu, the meet, the command line, the test scenarios, the stadium and the code. It then acts on seven playtest notes: Enter starting the game from any menu row, a menu athlete that matches the chosen event, hurdles cleared with a real jump, a run-up camera that shows the line ahead, a run-up that stops while the angle is set, a foul for running over the line after a throw, and better sound at launch.
Run details
Event types
- Thoughts123
- Messages37
- Tool calls168
- Commands56
- Reads57
- Writes54
- Done1