Manipulation · Door¶
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✗ 1h 00m | 09-28 16:19 | 1h 00m | — | 24.2M / 238k | $0.419 | deterministic 1, grader_error 0, n_actions 1000, video_rendered 1 |
| L✗ 1h 00m | 10-01 14:17 | 1h 00m | — | 30.0M / 171k | $0.433 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 405, 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 02:40 | 1h 00m | 124 | 15.3M / 109k | $3.09 | $3.21 | deterministic 1, grader_error 0, n_actions 1000, video_rendered 1 |
| L✗ 1h 00m | 09-30 22:45 | 1h 00m | 121 | 12.7M / 117k | $3.38 | $3.63 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 1000, replay_success 0, video_rendered 1 |
Claude Code 2.1.283 + Claude Opus 5.5, reasoning medium (OpenRouter)
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Grade |
|---|---|---|---|---|---|---|
| U✓ 49m | 09-28 19:45 | 49m | — | 5.8M / 55k | $4.81 | deterministic 1, grader_error 0, n_actions 1000, video_rendered 1 |
| L✗ 14m ⓘ | 09-28 17:33 | 14m | — | 4.4M / 55k | $4.07 | deterministic 1, grader_error 0, live_success 0, n_actions 1000, replay_success 0, video_rendered 1 |
Codex CLI 0.157.0 + GPT-6 Sol, reasoning medium (OpenRouter)closed
| Trial | Started | Agent time | Model requests | Tokens in / out | Est. cost | Grade |
|---|---|---|---|---|---|---|
| Unot run | — | — | — | — / — | — | — |
| L✗ 36m ⓘ | 09-28 17:51 | 36m | — | 8.3M / 66k | $2.60 | deterministic 1, grader_error 0, live_success 0, n_actions 1000, replay_success 0, video_rendered 1 |
Task instruction (upstream)
Walk to the door, press the handle, pull the door open towards the robot, and walk through it.
What HumanoidBench states about this task
| Success criteria | 1. the summed per-step reward over one episode reaches 600, HumanoidBench's own success bar (a total of rewards, not a number of steps; an episode is at most 1000 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 | h1-door-v0 |
| Robot | Unitree H1 (19 actuators) |
| Category | Manipulation |
| Capability Class | C4 · navigation and passage |
| Role | scored |
| Scoring | This one is a weighted sum rather than a product, so partial credit is real: getting the door open and getting the torso past the doorway are the two large terms, with reaching for the handle, moving the handle, and posture as small ones. The bar cannot be reached without opening the door. |
| Ends Early | The episode ends early if the pelvis drops too low. |
| Zero Action Return | 30.86 |
| Action Dim | 19 |
| 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 door's or the handle's angle, 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¶
The door opens towards the robot on a spring-loaded hinge: the robot must press the handle, pull the door open and step past the door leaf, while the stable crouches it finds tend to lean on the door and push it shut.
Capability notes¶
Not yet written.
Oracle demo review¶
No demo.
Discussion¶
Hard in both modes: unlimited mode unlocks the handle but opens the door only 26.5°, and limited mode never finds the handle. (@williamzhangNU)