Manipulation · Insert normal¶
Excluded from the benchmark
its bar does not measure the task: it can be reached without inserting (robot_coding_bench 2026-10-04)
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 18:48 | 1h 00m | — | 28.4M / 232k | $0.458 | deterministic 1, grader_error 0, n_actions 500, video_rendered 1 |
| L✗ 1h 00m | 10-01 14:58 | 1h 00m | — | 32.3M / 139k | $0.411 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 55, 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 04:50 | 1h 00m | 125 | 14.2M / 117k | $3.00 | $3.11 | deterministic 1, grader_error 0, n_actions 500, video_rendered 1 |
| L✗ 1h 00m | 09-30 23:24 | 1h 00m | 133 | 13.6M / 96k | $2.70 | $2.80 | deterministic 1, grader_error 0, live_success 0, missing_log 0, n_actions 35, 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)
A bar and two small open sockets lie on the table in front of the robot. Fit each end of the bar into the socket of the matching colour, then lift the assembly to about 1.1 m and hold it there, without dropping anything.
What HumanoidBench states about this task
| Success criteria | 1. the summed per-step reward over one episode reaches 350, 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-insert_normal-v0 |
| Robot | Unitree H1 with two Shadow hands (61 actuators) |
| Category | Manipulation |
| Capability Class | C7 · in-hand precision |
| Role | scored |
| Scoring | The step is a product of two groups: one covering posture and how close each end of the bar is to its socket, the other covering how high the sockets are and whether a hand is at a socket (the left hand at the dark one, or the right hand at the light one). Because the groups multiply, doing one of them well is worth little while the other is near zero — both have to come along together. |
| Ends Early | The episode ends the moment the bar or either socket drops below a height, which a dropped part does immediately, or the pelvis does. |
| Zero Action Return | 2.51 |
| 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 poses of the bar and the sockets, 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¶
Each end of a bar goes into the socket of its colour, then the robot lifts the assembly. The two groups of terms multiply, so getting only one of them right earns almost nothing.
Capability notes¶
Not yet written.
Oracle demo review¶
No demo.
Discussion¶
Blocked on the bar: unlimited mode lifts and turns both sockets but leaves the bar on the table; limited mode's grasps slip. (@williamzhangNU)