Feature #2243
Fires on “ best” (not labeled yet: read the examples)
- Fires on
- 0.41% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 59%
- agent 28%
- chat 12%
- python 1%
- worksheet 0%
Makes the model say
at friend friends to aboutyd if
Makes it not say
testsive] "!"Timop next belong
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 7.93only tell short stories and add numbers.<|assistant_end|><|bos|>Lila and Sam were best friends. They liked
-
story 7.77and they had many fun adventures together.<|assistant_end|><|bos|>Ben and Sam were best friends. They liked
-
story 7.57a happy cat named Kitty. Kitty lived in a small house with her best friend, a little
-
story 7.56there was a girl named Lily and a boy named Tom. They were best friends. They loved
-
story 7.51so much fun moving and playing on the farm. Molly and Billy became best friends, and they
-
story 7.43how to tell stories and add numbers.<|assistant_end|><|bos|>Anna and Lily were best friends. They liked
-
story 7.4, away from the wet water. They had lots of fun and became best friends. And they
-
story 7.37how to tell stories and add numbers.<|assistant_end|><|bos|>Lily and Ben were best friends. They liked
-
story 7.35a little dog named Max. Max lived in a small house with his best friend, a girl
-
story 7.31wanted to be friends with the unusual sheep. They played together and became best friends. And they
-
story 7.26the best eggs ever.⏎The end.<|bos|>Lily and Ben were best friends. They liked
-
story 7.2537 = 91540<|bos|>Tom and Mia were best friends. They lived