v1.3
Pretrained models for several games. Note that performance can vary greatly between runs on some games (particularly hard exploration games such as Frostbite, H.E.R.O. and Montezuma's Revenge). Reported scores achieved for all listed games except H.E.R.O. and Montezuma's Revenge.
Asteroids
| Reward | Q-values |
|---|---|
 |
 |
Boxing
| Reward | Q-values |
|---|---|
 |
 |
Breakout
| Reward | Q-values |
|---|---|
 |
 |
Beam Rider
| Reward | Q-values |
|---|---|
 |
 |
Enduro
| Reward | Q-values |
|---|---|
 |
 |
Freeway
| Reward | Q-values |
|---|---|
 |
 |
Frostbite
| Reward | Q-values |
|---|---|
 |
 |
H.E.R.O.
| Reward | Q-values |
|---|---|
 |
 |
Montezuma's Revenge
| Reward | Q-values |
|---|---|
 |
 |
Ms. Pac-Man
| Reward | Q-values |
|---|---|
 |
 |
Pong
| Reward | Q-values |
|---|---|
 |
 |
Q*bert
| Reward | Q-values |
|---|---|
 |
 |
Seaquest
| Reward | Q-values |
|---|---|
 |
 |
Space Invaders
| Reward | Q-values |
|---|---|
 |
 |
Video Pinball
| Reward | Q-values |
|---|---|
 |
 |