Feature #320
Fires on “ Mia” (not labeled yet: read the examples)
- Fires on
- 0.84% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 76%
- chat 20%
- worksheet 2%
- agent 1%
- python 1%
Makes the model say
loved wanted liked was alwaysERined likes
Makes it not say
But butfact Butankt more call
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 4.933 = 82.<|bos|>One day, a little girl named Sue went to the park
-
story 4.89stories and add numbers.<|assistant_end|><|bos|>One day, a little girl named Sue found a bag of
-
story 4.83answer is 8.<|assistant_end|><|bos|>One day, a little girl named Lily was playing with her
-
story 4.77is 53.<|assistant_end|><|bos|>One day, a little girl named Mia went to the kitchen
-
story 4.752347 apples.<|bos|>One day, a little girl named Mia went for a walk
-
story 4.74537.<|assistant_end|><|bos|>One day, a lively girl named Mia found a kay
-
story 4.72vent further danger.<|bos|>One day, a little girl named Amy went to the park
-
story 4.72⏎```<|assistant_end|><|bos|>One day, a girl named Mia found an orange toy
-
story 4.72is 25.<|assistant_end|><|bos|>One day, a little girl named Lily was troubled. She
-
story 4.711021.<|bos|>One day, a little girl named Amy was sad. She
-
story 4.7the sun began to set.<|bos|>One day, a little girl named Sue went to the park
-
story 4.71520.<|bos|>One day, a little girl named Lily went to the park