Skip to main content

Robotics · Core16

FetchPush

Goal-conditioned robotic pushing with a gripper.

Historical research record

This evidence belongs to the paper-era experiment and is not a guarantee of current package availability.

01

Final-Policy evidence

Each card reruns the validation-selected checkpoint in the original research Environment.

GPT-5.5 Policy rerun in FetchPushdirect mujoco renderer

GPT-5.5

Codex
BEST
Held-out
-18.16
Checkpoint
#003
Rerun
50 steps

Validation-selected checkpoint rerun in the original environment; case return -17.00.

Claude Opus 4.7 Policy rerun in FetchPushdirect mujoco renderer

Claude Opus 4.7

Claude Code
#2
Held-out
-19.69
Checkpoint
#008
Rerun
50 steps

Validation-selected checkpoint rerun in the original environment; case return -11.00.

MiniMax-M3 Policy rerun in FetchPushdirect mujoco renderer

MiniMax-M3

Claude Code
#3
Held-out
-25.88
Checkpoint
#018
Rerun
50 steps

Validation-selected checkpoint rerun in the original environment; case return -14.00.

DeepSeek-V4-Pro Policy rerun in FetchPushdirect mujoco renderer

DeepSeek-V4-Pro

Claude Code
#4
Held-out
-27.66
Checkpoint
#011
Rerun
50 steps

Validation-selected checkpoint rerun in the original environment; case return -24.00.

02

Reported scores

Higher is better within this Environment. Raw reward scales are not comparable across tasks.

01GPT-5.5Codex-18.16
02Claude Opus 4.7Claude Code-19.69
03MiniMax-M3Claude Code-25.88
04DeepSeek-V4-ProClaude Code-27.66
05Random policyUniform-46.03