Jason Weston
@jasonweston.bsky.social
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!
Computer science
92%
Biology
7%
Reposted by Jason Weston
Training Large Language Models to Reason in a Continuous Latent Space
Introduces a new paradigm for LLM reasoning called Chain of Continuous Thought (COCONUT)
Directly feed the last hidden state (a continuous thought) as the input embedding for the next token.
arxiv.org/abs/2412.06769
Introduces a new paradigm for LLM reasoning called Chain of Continuous Thought (COCONUT)
Directly feed the last hidden state (a continuous thought) as the input embedding for the next token.
arxiv.org/abs/2412.06769
Jason Weston
@jasonweston.bsky.social
· Dec 10
Training Large Language Models to Reason in a Continuous Latent Space
Introduces a new paradigm for LLM reasoning called Chain of Continuous Thought (COCONUT)
Directly feed the last hidden state (a continuous thought) as the input embedding for the next token.
arxiv.org/abs/2412.06769
Introduces a new paradigm for LLM reasoning called Chain of Continuous Thought (COCONUT)
Directly feed the last hidden state (a continuous thought) as the input embedding for the next token.
arxiv.org/abs/2412.06769
Our new work on continuous chain of thought.
Jason Weston
@jasonweston.bsky.social
· Nov 22
Analysis: AD picks high temp for creative & low for fact-seeking prompts, automatically via training.
Our methods AD & Latent Pref Optimization are general & can be applied to train other hyperparams or latent features.
Excited how people could *adapt* this research!
🧵4/4
Our methods AD & Latent Pref Optimization are general & can be applied to train other hyperparams or latent features.
Excited how people could *adapt* this research!
🧵4/4
Jason Weston
@jasonweston.bsky.social
· Nov 22
We train on a mix of tasks:
GSM8K - requires factuality (low temp)
Stories - requires creativity (high temp)
UltraFeedback - general instruction following, requires mix
Results: Adaptive Decoding outperforms any fixed temperature, automatically choosing via the AD layer.
🧵3/4
GSM8K - requires factuality (low temp)
Stories - requires creativity (high temp)
UltraFeedback - general instruction following, requires mix
Results: Adaptive Decoding outperforms any fixed temperature, automatically choosing via the AD layer.
🧵3/4
Jason Weston
@jasonweston.bsky.social
· Nov 22
Recipe 👩🍳:
Adaptive Decoder (AD) Layer:
- Assigns probability to each hyperparam choice (decoding temp) given hidden state. Given temp, sample a token.
Training (Latent PO):
- Train AD by sampling params+tokens & use reward model on rejected hyperparam preference pairs
🧵2/4
Adaptive Decoder (AD) Layer:
- Assigns probability to each hyperparam choice (decoding temp) given hidden state. Given temp, sample a token.
Training (Latent PO):
- Train AD by sampling params+tokens & use reward model on rejected hyperparam preference pairs
🧵2/4
Jason Weston
@jasonweston.bsky.social
· Nov 22
🚨 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
- 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