Feature #1847
Fires on “ friends” (not labeled yet: read the examples)
- Fires on
- 0.28% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 78%
- chat 13%
- worksheet 6%
- agent 2%
- python 1%
Makes the model say
and foreveris news instead drank6
Makes it not say
Mom<|think_start|> whether theofger Mom edit
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 7.83saved the bird. The bird, Tim, and the car all became friends and played together.
-
story 7.77was not stuck anymore and was very happy. Spot and the cat became friends and played together every
-
story 7.69and surf too. Tim and the fish became friends and they surfed
-
story 7.65bird keep the spear. The king and the little bird became good friends, and they lived
-
story 7.58mattress. The mattress was happy now. They became friends and played together every
-
story 7.42silly and strong he was.⏎The little girl and the monster became friends and watched cartoons
-
chat 7.37free. The bird was happy. The compassionate man and the bird became friends. They played together
-
chat 7.35. They could now eat the broken cake. Tim and the birds became friends and shared the cake
-
story 7.31. Let's be friends." So, the plane and the fish became friends and played together every
-
chat 7.3. The mouse was scared, but the cat was nice. They became friends and played together.
-
chat 7.29free. The bird was happy. The compassionate man and the bird became friends. They played together
-
story 7.29I know you didn't mean it." The deer and the bug became friends. They played together