Feature #452
Fires on “ tree” (not labeled yet: read the examples)
- Fires on
- 0.54% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 34%
- worksheet 32%
- chat 29%
- agent 5%
- python 0%
Makes the model say
9 now prelude and =ance1ue
Makes it not say
C( bigLab only hardly
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 1.9under a big tree and shared the nuts
-
story 1.88climbed to the top of a tall tree. He put on
-
story 1.81text.py`, line 1.<|assistant_end|><|bos|>Once upon a time, a bald man named
-
story 1.78to ten, and Ben and Max run to hide behind a big tree.⏎"Read
-
story 1.77As Tim rode his bike, he saw a smelly cat stuck in a tree. "Don't
-
story 1.76: `add(a, b)`.<|assistant_end|><|bos|>Once upon a time, there was a boy
-
chat 1.74for a walk. The dog saw a small cat sitting under a tree.<|assistant_end|><|user_start|>Bye
-
story 1.73rock friend.⏎Suddenly, Tim saw a small cat stuck in a tree. The cat was
-
story 1.73is_divisible(a, b)`.<|assistant_end|><|bos|>Once upon a time, there was a big
-
story 1.72<|assistant_start|>Created `greeting.py`.<|assistant_end|><|bos|>Once upon a time, there was a little
-
story 1.72<|assistant_start|>Created `greeting.py`.<|assistant_end|><|bos|>Once upon a time, a little girl named
-
story 1.71<|tool_end|><|assistant_start|>Created `hello.py`.<|assistant_end|><|bos|>Once upon a time, there was a girl