Feature #3799
Fires on “ only” (not labeled yet: read the examples)
- Fires on
- 0.21% of tokens
- Mostly in
- chat
- Layer
- after block 4
Where it fires
- chat 47%
- story 26%
- agent 12%
- worksheet 8%
- python 7%
Makes the model say
ly mini)`. just tellcaseidereplaceAll
Makes it not say
The theThe is was, equals sum
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 3.51space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.45for?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.35space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.31space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.3space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.3space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.3for?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.29cook pasta<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.28end?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.28code?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.27space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
-
chat 3.27in?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and