Feature #901
Fires on “ was” (not labeled yet: read the examples)
- Fires on
- 0.42% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 80%
- chat 12%
- agent 5%
- python 2%
- worksheet 1%
Makes the model say
to,i import."t," from
Makes it not say
*calculatorpath<|think_end|> command listoldString
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 5.36was big and yummy.
-
story 5.36was still alert. Then
-
story 5.36was a little dog named
-
chat 5.36was very hungry. He
-
story 5.36was there to save the
-
story 5.36was a small dog named
-
story 5.36was very happy he found
-
story 5.36was glad and started playing
-
chat 5.36was a sailor.<|assistant_end|>
-
chat 5.36was a hot day.
-
story 5.36was red and blue.
-
story 5.36was so lovely and strong