Feature #1110
Fires on “ are” (not labeled yet: read the examples)
- Fires on
- 0.37% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 70%
- agent 14%
- chat 11%
- python 3%
- worksheet 2%
Makes the model say
sad so was faces not?" friends were
Makes it not say
aei]`ten],ge Letitems i
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 3.28why are you being so
-
story 3.17"Why are you sad, little boy
-
story 3.14and went to talk to him. She asked, "Ben, why are you not playing with
-
story 3.05bird. The bird said, "Hi, Will! Why are you so excite?" W
-
story 3.04ine loved to question things. It would ask the trees, "Why are you so tall?"
-
story 2.99er marched to the cat. He asked, "Why do you look guilty?"
-
story 2.98, who looked tired and weak.⏎"Ben, why do you look so tired?" asked
-
story 2.97Why are you sad?"
-
story 2.94doing?" he says. "Why are you fighting over a car
-
story 2.94went to the dragon's cave. She asked the dragon, "Why are you so noisy?"
-
story 2.9girl came to visit the cat. She asked the cat, "Why are you so grumpy?"
-
story 2.9John and became angry. He shouted at John and asked him why he was recording him