Feature #3341
Fires on “ They” (not labeled yet: read the examples)
- Fires on
- 0.70% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 71%
- agent 12%
- chat 11%
- worksheet 4%
- python 2%
Makes the model say
everywhere alwaysread yours liked like[-ave
Makes it not say
+'s3 werepy the12
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 2.86answer is 78.<|assistant_end|><|bos|>Tom and Sam are friends. They like to play with
-
story 2.8590<|think_end|>790<|assistant_end|><|bos|>Lily and Ben are friends. They like to play in
-
story 2.7692725.<|bos|>Lily and Tom are friends. They like to play outside
-
story 2.75, there was a little dog named Spot. Spot was a dependable dog who loved to play outside
-
story 2.74key. She took it everywhere she went. Mia was an independent girl who liked to do things
-
story 2.67, there was a little boy named Tim. Tim was a normal boy who liked to play with
-
chat 2.66a little dog named Tim. Tim was an obedient dog who always listened to his
-
story 2.66, there was a little boy named Tom. Tom was an adventurous boy who loved to play outside
-
story 2.631706.<|assistant_end|><|bos|>Tim and Sam are friends. They like to play with
-
story 2.58is 2603.<|bos|>Anna and Ben are friends. They like to play pretend
-
story 2.57cactus was now their new friend.<|bos|>Sara and Ben are twins. They like to play with
-
story 2.56item` in this project.<|assistant_end|><|bos|>Anna and Ben are friends. They like to play in