Feature #3869
Fires on “ everyone” (not labeled yet: read the examples)
- Fires on
- 0.70% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 75%
- agent 16%
- chat 6%
- python 2%
- worksheet 1%
Makes the model say
sad happy feelOnceakes angry smile special
Makes it not say
"!"`.()`. He]`. But round "-
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 6.96growing, and Sue and Tim learned that sharing with others can make everyone happy.<|bos|>Once
-
story 6.52felt very happy. The moral of the story is that sharing can make everyone happy, even when
-
story 6.25happy and thanked Lily.⏎The two friends learned that sharing can make everyone happy and safe.
-
story 6.01three friends played together every day. They learned that helping others can make you happy and bring you
-
story 5.79fun. Tim and his friends laughed. They learned that sharing can make others happy too. And
-
story 5.73played together all day. Tim learned that sharing and helping others can make you feel good too.
-
story 5.5please don't kick my bed!"⏎Tom was sad that he made Lily upset. He said
-
story 5.41fight. And Lucy learned that asking for help and being kind can make everyone happy.<|bos|><|system_start|>
-
story 5.38feel good. The moral of the story is that helping others can make you happy too.<|bos|>
-
story 5.23fun. The moral of the story is that sharing fun with others makes everyone happy.<|bos|><|system_start|>
-
story 5.2She said, "Let's go play together then! That will make you feel better!"⏎
-
story 5.2Her mom was happy too, because she knew that the ball would make Lucy happy. The thick