Feature #22
Fires on “ just” (not labeled yet: read the examples)
- Fires on
- 0.70% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 66%
- agent 20%
- chat 10%
- python 2%
- worksheet 2%
Makes the model say
because needed wasoldString of and right there
Makes it not say
⏎ked They1 =29it
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 3.09just like the boat.
-
story 3.09just a toy I use
-
story 2.79's okay, it's okay, don't be afraid. It was just your shadows.
-
story 2.48even more fun.<|bos|>
-
story 2.34big cow laughed and said, "That's not real! It's just a toy!" They
-
chat 2.32<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story, or some numbers to
-
chat 2.29<|assistant_end|><|user_start|>Hi<|user_end|><|assistant_start|>Hi there! Would you like a story, or some numbers to
-
chat 2.26then, something unexpected happened. The fire was not real. It was just a toy fire.
-
chat 2.24<|user_start|>Hi!<|user_end|><|assistant_start|>Hi there! Would you like a story, or some numbers to
-
chat 2.24hey there<|user_end|><|assistant_start|>Hi there! Would you like a story, or some numbers to
-
chat 2.2<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story, or some numbers to
-
story 2.2it was beautiful and Abig