NextLat is going to NeurIPS ๐
Next-token prediction is myopic. What if transformers learn to predict their own next latent state? ๐ We present ๐ก๐ฒ๐ ๐-๐๐ฎ๐๐ฒ๐ป๐ ๐ฃ๐ฟ๐ฒ๐ฑ๐ถ๐ฐ๐๐ถ๐ผ๐ป (๐ก๐ฒ๐ ๐๐๐ฎ๐): a self-supervised learning method that teaches transformers to form compact world models for reasoningโฆ
Not enough discussion yet.
No sentiment analysis available yet.
Not enough discussion yet.
No sentiment analysis available yet.
โ
Not ranked yet
โ
Not ranked yet
