Sign in

Jason Weston

@jasonweston.bsky.social
585 followers 342 following 5 posts

Senior Director, Research Scientist @ Meta FAIR + Visiting Prof @ NYU. Pretrain+SFT: NLP from Scratch (2011). Multilayer attention+position encode+LLM: MemNet (2015). Recent (2024): Self-Rewarding LLMs & more!

PostsRepliesMedia
Jason Weston @jasonweston.bsky.social · 10/12/2024
Our new work on continuous chain of thought.
040
Jason Weston @jasonweston.bsky.social · 22/11/2024
🚨 Adaptive Decoding via Latent Preference Optimization 🚨 - New layer for Transformer, selects decoding params automatically *per token* - Learnt via new method Latent Preference Optimization - Outperforms any fixed temperature decoding, choosing creativity or factuality arxiv.org/abs/2411.09661 🧵1/4
2436