Preprint
Oct 2026
Complementary Future Token Pruning for World Action Models
The key insight is that future visual predictions support action generation and therefore need not preserve every visual detail required for high-fidelity video synthesis, and CF-Prune, a training-free framework for pruning future-frame tokens in pretrained WAMs, is proposed.
Baekseung Kim, Junhoo Lee, Sohyeon Hong et al.
· 0 citations