Hold Cup¶
Agent runs
Each mode of this task runs once per run. Est. cost is tokens at list price (data/prices.yml), never a bill; see the RoboWits runs for every task of a run.
Codex + GPT-6 Luna, reasoning xhigh (ChatGPT login)default run
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Grade |
|---|---|---|---|---|---|---|
| U✓ 1h 15m | 09-26 03:39 | 1h 15m | 293 | 36.2M / 145k | $0.495 | deterministic 1, grader_error 0, instance_exact 1, instance_ok 1, n_actions 1073, video_rendered 1 |
| L✓ 1h 30m | 09-29 17:44 | 1h 30m | 229 | 29.1M / 176k | $0.438 | deterministic 1, grader_error 0, live_success 1, n_actions 1795, replay_success 1, video_rendered 1 |
Task instruction (upstream)
Put the cup on the green target area on the slope without sliding down.
Playback speed
What RoboWits states about this task
| Success criteria | 1. Cup's 2D footprint overlaps target area by at least 50% 2. Cup is stable with near-zero linear velocity 3. Cup is upright (tilt from vertical <= 20 degrees) 4. Cup remains within table bounds 5. Cup is not being held by either gripper |
| Objects | coaster cup slope target area |
| Physics | rigid |
| Episode budget | 200 control steps |
| Default control mode | EE_ABS |
| Evaluation scenes | 50 |
| Published mutations | 5 |
| BlenderKit assets | 2 (not redistributable — mounted from the host) |
| Registry id | robowits/09-hold-cup-v0 |
| Environment class | HoldCupEnv |
| Defined in | gs_gym/envs/robowits/09_hold_cup.py |
From https://github.com/UMass-Embodied-AGI/RoboWits @ 9cc30ae.
Tags¶
Task DomainManipulation
Why this task is interesting¶
Not yet written.
Capability notes¶
Not yet written.
Oracle demo review¶
Not yet reviewed.