Feature #3012
Fires on “ to” (not labeled yet: read the examples)
- Fires on
- 0.36% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 83%
- chat 11%
- agent 3%
- worksheet 2%
- python 0%
Makes the model say
mad cryus it we smile us cat
Makes it not say
+'t never")list modelaeiou⏎⏎
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 6.01looked everywhere, but he could not find his lost treasure. He started to feel very upset and
-
story 5.99go to the party!"⏎Joe was sad now and he started to cry. His mommy
-
story 5.98out, but he couldn't. He felt helpless. He started to cry.⏎Lily
-
story 5.96then, the sky got dark and it started to rain. Sue began to worry. She didn
-
story 5.95, but she didn't have any money. She was sad and began to weep.⏎
-
story 5.86would come down, but nothing happened. She got very sad and started to cry. But suddenly
-
story 5.83identally pinched the rabbit! The rabbit was very scared and started to whimper.
-
story 5.66could not find any food. Tom started to worry. He thought
-
story 5.66Boo!"⏎Sue got very scared and started to cry. She dropped
-
story 5.6does not know where she is. She is scared. She starts to cry. She call
-
story 5.58while playing with the bottle, it broke. Tim got hurt and began to cry. His mom
-
story 5.58⏎Finally, Joe got stuck. He was so sad that he started to cry. But then