Feature #1091
Fires on “ boy” (not labeled yet: read the examples)
- Fires on
- 0.57% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 61%
- chat 20%
- worksheet 9%
- python 8%
- agent 1%
Makes the model say
Lily girl Sue go his boy work sat
Makes it not say
newString))SoredreplaceAll huggedlloSet
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 3.69did not know what to do.⏎Then, they saw a man. He had a ladder
-
story 3.43⏎2 plus 1 equals 3.<|bos|>One day, a boy named Tim found a
-
story 3.35The answer is 638.<|assistant_end|><|bos|>One day, a boy named Tom was very
-
story 3.32help her mom and be a careful girl.<|bos|>One day, a boy named Tim saw a
-
story 3.27boy named Tim. He
-
story 3.27boy. What do you
-
story 3.27boy named Tim. Tim
-
story 3.27boy comes to the garden
-
story 3.27boy said, "Hi
-
story 3.27boy smiled and said,
-
story 3.26843.<|assistant_end|><|bos|>One day, a boy named Tim went to
-
story 3.24never eat dinner in the tower again!<|assistant_end|><|bos|>One day, a boy named Tim went for