native env renderGPT-5.5
Codex- Held-out
- 95.22
- Checkpoint
- #005
- Rerun
- 88 steps
Validation-selected checkpoint rerun in the original environment; case return 95.69.
Control · Core16
ContinuousCar
Continuous throttle control with delayed goal reward.
Historical research record
This evidence belongs to the paper-era experiment and is not a guarantee of current package availability.
Final-Policy evidence
Each card reruns the validation-selected checkpoint in the original research Environment.
native env render#2
native env renderClaude Opus 4.7
Claude Code- Held-out
- 98.77
- Checkpoint
- #010
- Rerun
- 344 steps
Validation-selected checkpoint rerun in the original environment; case return 98.89.
native env renderMiniMax-M3
Claude Code- Held-out
- 91.88
- Checkpoint
- #010
- Rerun
- 143 steps
Validation-selected checkpoint rerun in the original environment; case return 91.61.
native env renderDeepSeek-V4-Pro
Claude Code- Held-out
- 94.48
- Checkpoint
- #012
- Rerun
- 69 steps
Validation-selected checkpoint rerun in the original environment; case return 94.50.
Reported scores
Higher is better within this Environment. Raw reward scales are not comparable across tasks.
01Claude Opus 4.7Claude Code
98.7702GPT-5.5Codex
95.2203DeepSeek-V4-ProClaude Code
94.4804MiniMax-M3Claude Code
91.8805Random policyUniform
-33.36