osazuwa/3D-dungeon-crawler-video

3D Dungeon Crawler Video

Deterministic 23-second Unity episodes for observational world-model training. The 10,000 episodes are split into 8,000 train, 1,000 validation, and 1,000 test episodes.

Unity renders at 512x288 for supersampling; each clips/*.mp4 is area-downscaled and stored at 256x144 and 30 fps. Training samples every third frame, yielding 230 frames and an 18x32 tokenizer grid. Matching arrays/*.npz files contain dag_ticks, action_tokens, and final_state.

All episodes are observational (do={}), the curse variable Z is visible, and splits are deterministically balanced across all nine binary causal variables: D, Z, V, W, X, M, A, Y, R.