Texonom
Texonom/
Engineering
Engineering/Data Engineering/Artificial Intelligence/Machine Learning/Neural Network/Neural Network Structure/Seq2Seq/Attention Mechanism/Reversing Transformer/
Transformer Feed-Forward Layer
Search

Transformer Feed-Forward Layer

Creator
Creator
Seonglae ChoSeonglae Cho
Created
Created
2024 Apr 18 16:36
Editor
Editor
Seonglae ChoSeonglae Cho
Edited
Edited
2024 Apr 19 14:34
Refs
Refs
Neuron SAE
https://arxiv.org/pdf/2203.14680.pdf
  • A FFN subvalue can help increase probabilities of tokens with largest logits.
  • it can reduce probabilities of tokens with smallest logits.
  • It can distinguish two tokens with different logits
There are tens of thousands of token pairs in a FFN subvalue, so one FFN subvalue can distinguish many tokens. Last, a FFN subvalue can be a “query” to activate other FFN subvalues.
 
 
Exploring the Residual Stream of Transformers for Mechanistic Interpretability — Explained
— by Zeping Yu, Dec 24, 2023
Exploring the Residual Stream of Transformers for Mechanistic Interpretability — Explained
https://medium.com/@zepingyu/123-cb62513f5d50
Exploring the Residual Stream of Transformers for Mechanistic Interpretability — Explained

Transformer Feed-Forward Layers Are Key-Value Memories

aclanthology.org
https://aclanthology.org/2021.emnlp-main.446.pdf

Factual GPT before mono-semanticity space autoencoder (some limitations)

arxiv.org
https://arxiv.org/pdf/2202.05262.pdf
 
 
 

Recommendations

Texonom
Texonom/
Engineering
Engineering/Data Engineering/Artificial Intelligence/Machine Learning/Neural Network/Neural Network Structure/Seq2Seq/Attention Mechanism/Reversing Transformer/
Transformer Feed-Forward Layer