Feature #1804
Fires on “ you” (not labeled yet: read the examples)
- Fires on
- 0.53% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 44%
- agent 40%
- chat 14%
- python 2%
- worksheet 1%
Makes the model say
?"!"24 okay<|user_end|> my3
Makes it not say
they` Theylren theirine
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 4.07scare you. It was a joke. We are sorry. Are you mad at us?"
-
story 4.02earthquake stopped and everything was calm again.⏎"Are you okay, Spot?"
-
story 3.83, "Hello! Why are you ringing that bell
-
story 3.75says, "Okay, Lily. It's our turn. Are you ready?"⏎Lily
-
story 3.59you can make lots of
-
agent 3.59you?<|user_end|><|assistant_start|>I
-
chat 3.59you tell me what
-
worksheet 3.59you get 12
-
story 3.59you help me?" Mia
-
agent 3.59you run run.py
-
story 3.59you too, Tom.
-
chat 3.59you tell me a story