Ethan Ewer

[Kangwook Lee @Kangwook_Lee]

Well, that's it!

While ENTP might not replace decoder models for now, it offers a promising direction if we overcome current computational limitations.

Check out our full paper on Arxiv for details—we’d love your feedback!
https://t.co/mAEM2aT2vV

[Kangwook Lee @Kangwook_Lee]

P.S.: ENTP also happens to be my Myers-Briggs type—pure coincidence, I swear! 😅

[Kangwook Lee @Kangwook_Lee]

🚀 Excited to share our work on Encoder-only Next Token Prediction (ENTP)!

While most successful LLMs are decoder-based, we asked: Can encoder-only TFs be used for next-token prediction?

Yes!

Moreover, ENTP might be better than decoder-only models!!! 😎