Dynamic3D · Shelf¶
Excluded from the benchmark
not among the hardest (robot_coding_bench 2026-10-04): GPT-6 Luna passed it in unlimited mode in 21 min
Greyed and last in the task list, left out of the benchmark's numbers; its demo and agent runs stay (excluded: in state/tasks/kinder.yml).
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 KinDER runs for every task of a run.
Codex CLI 0.157–0.159 + GPT-6 Luna, reasoning medium (OpenRouter)default run
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Grade |
|---|---|---|---|---|---|---|
| U✓ 21m | 09-29 00:02 | 21m | — | 10.3M / 87k | $0.161 | deterministic 1, grader_error 0, missing_trajectory 0, n_actions 955, video_rendered 1 |
| L✗ 1h 00m | 10-01 15:23 | 1h 00m | — | 32.3M / 166k | $0.428 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 305, replay_success 0, video_rendered 1 |
1 other run(s) of KinDER
Codex CLI 0.159.2 + GPT-6.1 Sol, reasoning medium (OpenRouter)
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Billed | Grade |
|---|---|---|---|---|---|---|---|
| U✓ 6m | 10-01 05:53 | 6m | 24 | 991k / 8k | $0.294 | $0.324 | deterministic 1, grader_error 0, missing_trajectory 0, n_actions 785, video_rendered 1 |
| L✓ 15m | 10-01 02:42 | 15m | 70 | 3.0M / 14k | $0.581 | $0.619 | deterministic 1, grader_error 0, live_success 1, missing_log 0, n_actions 784, replay_success 1, video_rendered 1 |
Task instruction (upstream)
Pick up the dark-red cube and the dark-green cube from the floor and put both in the top compartment of the brown cupboard, which has three compartments one above another. The task succeeds the moment both cubes rest on that compartment's floor.
What KinDER states about this task
| Success criteria | 1. KinDER's own goal check (the environment's terminated) fires within the episode: at most 1000 control steps (100 s at 10 Hz); no partial credit2. unlimited: both fresh-process replays of the handed-in trajectory reach it and end in the same state 3. limited: the one episode reaches it (no reset in our runs); the service records the episode, and it replays to the same state |
| Env Id | kinder/Shelf3D-o2-v0 |
| Robot | TidyBot++: holonomic base, Kinova Gen3 7-DoF arm, Robotiq 2F-85 gripper |
| Category | Dynamic3D (MuJoCo) |
| Instance | seed 0: the scene KinDER builds from it, pinned by its digest |
| Holding Still | does not reach the goal (checked by running an episode of zero actions) |
| Action Dim | 11 |
| Actions | base pos and yaw (3), arm joints (7), gripper pos (1); each base and joint delta at most 0.1 a step |
| State Dim | 61 |
| Limited Mode | base RGB and wrist RGB-D cameras (480×640), base odometry, the arm's joint angles and velocities, the fingers' position. Not where any object or the goal region is, not the goal check before the episode ends |
| Agent Budget | 3600 s of wall clock per mode |
From https://github.com/Princeton-Robot-Planning-and-Learning/kindergarden @ 5b2dbac, as defined in our task definitions @ 7d6a7a44.
Tags¶
Why this task is interesting¶
Two cubes on the floor go into the top of a cupboard's three compartments, resting on its floor. The robot has to find the cupboard and drive up to it, then reach in through the front: a vertical wrist hits the compartment's roof.
Capability notes¶
Not yet written.
Oracle demo review¶
No demo.
Discussion¶
Both modes pass: limited mode found the hidden red cube by turning the base and entered the top compartment with the gripper held horizontal. (@williamzhangNU)