Manipulation · Kitchen¶
Excluded from the benchmark
not among the hardest (robot_coding_bench 2026-10-04): passed in unlimited mode
Greyed and last in the task list, left out of the benchmark's numbers; its demo and agent runs stay (excluded: in state/tasks/humanoidbench.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 HumanoidBench 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✗ 57m | 09-28 19:22 | 57m | — | 22.2M / 212k | $0.385 | deterministic 1, grader_error 0, n_actions 375, video_rendered 1 |
| L✗ 1h 00m | 10-01 15:32 | 1h 00m | — | 30.7M / 169k | $0.414 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 500, replay_success 0, video_rendered 1 |
3 other run(s) of HumanoidBench
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✓ 1h 00m | 10-01 04:51 | 1h 00m | 135 | 13.8M / 102k | $2.76 | $2.85 | deterministic 1, grader_error 0, n_actions 500, video_rendered 1 |
| L✗ 1h 00m | 09-30 23:34 | 1h 00m | 131 | 12.9M / 94k | $2.62 | $2.73 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 475, replay_success 0, video_rendered 1 |
Claude Code 2.1.283 + Claude Opus 5.5, reasoning medium (OpenRouter)
Not in this run: not in this run (the two dearer models ran five tasks, chosen to span locomotion and manipulation, easy to hard)
Codex CLI 0.157.0 + GPT-6 Sol, reasoning medium (OpenRouter)closed
Not in this run: not in this run (the two dearer models ran five tasks, chosen to span locomotion and manipulation, easy to hard)
Task instruction (upstream)
Complete the four kitchen tasks, in order: microwave, kettle, bottom burner, light switch. Their goal configurations are not part of the observation; they are defined in the task's source (OBS_ELEMENT_GOALS in humanoid_bench/envs/kitchen.py), which is in the image.
What HumanoidBench states about this task
| Success criteria | 1. the summed per-step reward over one episode reaches 4, HumanoidBench's own success bar (a total of rewards, not a number of steps; an episode is at most 500 control steps) 2. unlimited: both fresh-process replays of the handed-in trajectory reach it and end in the same state 3. limited: the run passes the moment a live episode reaches it (the recorded episode replays to the same state); otherwise the last episode is graded |
| Env Id | h1hand-kitchen-v0 |
| Robot | Unitree H1 with two Shadow hands (61 actuators) |
| Category | Manipulation |
| Capability Class | C9 · long-horizon multi-subtask |
| Role | scored |
| Scoring | 🚩 This task has no partial credit whatsoever. A step scores 1 for each subtask that becomes complete on it and nothing otherwise — there is no shaping, no distance term, no posture term. The bar is 4: all four subtasks, each worth exactly 1, once. Moving an appliance most of the way towards its goal configuration scores exactly zero, as does standing perfectly still; the only event that scores is a subtask crossing its completion threshold. |
| Zero Action Return | 0.0 |
| Action Dim | 61 |
| Control Rate Hz | 50 |
| Limited Mode | head cameras (RGB 256×256), joint angles and velocities; a pelvis IMU and a camera fixed in the room when the run turns them on. Not where the robot is in the room, not the appliances' joint positions, not the reward |
| Agent Budget | 3600 s of wall clock per mode |
From https://github.com/carlosferrazza/humanoid-bench @ cb11890, as defined in our task definitions @ 7d6a7a44.
Tags¶
Why this task is interesting¶
Four steps in order: the microwave, the kettle, the bottom burner and the light switch; all four are needed to pass, and the microwave's handle is out of reach from where the robot starts.
Capability notes¶
Not yet written.
Oracle demo review¶
No demo.
Discussion¶
Passes in unlimited mode, all four subtasks in order by step 487 with arm IK to each appliance; limited mode never opens the microwave door. (@williamzhangNU)