Discrete diffusion models sample tokens independently at each step, losing uncertainty from earlier predictions.
Loopholing uses self-conditioning to pass a continuous state between steps but still samples at each step.
Could we go all the way with self-conditioning? 🤔
🧵1/7
Loopholing uses self-conditioning to pass a continuous state between steps but still samples at each step.
Could we go all the way with self-conditioning? 🤔
🧵1/7