Language steering
Reyn
Steer play style, movement, and combat with natural-language instructions.
- 26 characters
- Language input
Frisson Labs / Models
Explore our policies, language-steerable agents, and experiments in cooperative play.
Language steering
Steer play style, movement, and combat with natural-language instructions.
Experimental 2v2
Give each Fox teammate its own instruction, or send the same command to both.
Combo experiment
Two Fox agents work together to extend combos against a non-fighting dummy.
Fighting policies
Explore six fighting-policy checkpoints, from pretraining through supervised learning to reinforcement learning.
| Size | BasePretrained | ExpertSupervised | ArenaRL |
|---|---|---|---|
| 10M | 10M Base | 10M Expert | 10M Arena |
| 75M | 75M Base | 75M Expert | 75M Arena |
Arena continues Expert with reinforcement learning in Fox mirrors. The shared weights retain the 26-character interface.
| Configuration | 10M | 75M |
|---|---|---|
| Parameters | 10,163,629 | 75,305,709 |
| Transformer blocks | 5 | 11 |
| Hidden width | 384 | 768 |
| Query / KV heads | 6 / 2 | 12 / 3 |
| Context | 256 frames | 256 frames |