Best published lab score ✓ Verified
500,350 points
Privileged trajectory search · 0 model calls
Whole declared trace/input, two NEW native resets + result-bound pixel/round gates; third whole recording
The laboratory scoreboard
The best published scores, the replay behind each run and the evidence that checks it.
Scores are grouped by game and research method. Privileged trajectory search and automated policy search have their own conditions; these are laboratory milestones. Read the methodology →
Best published lab score ✓ Verified
Privileged trajectory search · 0 model calls
Whole declared trace/input, two NEW native resets + result-bound pixel/round gates; third whole recording
PleiadsBest published lab score ✓ Verified
Automated policy search · 0 model calls
Exact replay · 2,239/2,239 sampled frames matched
Every step counts
| Game | Score | Published | Method | Replay & evidence |
|---|---|---|---|---|
| Pengo | 500,350 | Privileged trajectory search0 model calls | Report · Evidence · Video ↗ | |
| Pengo | 301,030 | Privileged trajectory search0 model calls | Report · Evidence · Video ↗ | |
| Pengo | 216,480 | Privileged trajectory search0 model calls | Report · Evidence · Video ↗ | |
| Pengo | 100,400 | Privileged trajectory search0 model calls | Report · Evidence · Video ↗ | |
| Pengo | 31,040 | Privileged trajectory search0 model calls | Report · Evidence · Video ↗ | |
| Pengo | 21,210 | Privileged trajectory search0 model calls | Report · Evidence · Video ↗ | |
| Pleiads | 5,750 | Automated policy search0 model calls | Report · Evidence · Video ↗ | |
| Pleiads | 5,600 | Automated policy search0 model calls | Report · Evidence · Video ↗ | |
| Pleiads | 5,540 | Automated policy search0 model calls | Report · Evidence · Video ↗ |