Feature #2167
Fires on “ Tim” (not labeled yet: read the examples)
- Fires on
- 0.15% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 75%
- chat 20%
- agent 3%
- python 2%
- worksheet 0%
Makes the model say
Tim, or) - they+ was
Makes it not say
aTrueHelloerFalsenenBut
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 2.716.<|assistant_end|><|bos|>Once upon a time, there was a boy named Tim. He liked to
-
story 2.594.<|bos|>Once upon a time, there was a young boy named Tim. Tim loved to
-
story 2.57.<|assistant_end|><|bos|>Once upon a time, there was a little boy named Tim. Tim was a
-
story 2.56big fish surfing together.<|bos|>Once, there was a boy named Tim. Tim had a
-
story 2.55in.<|bos|>Once upon a time, there was a little boy named Tim. He liked to
-
story 2.55amazing.<|bos|>Once upon a time, there was a little boy named Tim. Tim loved to
-
story 2.549.<|bos|>Once upon a time, there was a little boy named Tim. He lived in
-
story 2.543.<|assistant_end|><|bos|>Once upon a time, there was a boy named Tim. Tim was a
-
chat 2.52.<|user_end|><|assistant_start|>Once upon a time, there was a little boy named Tim. Tim loved
-
story 2.51<|assistant_end|><|bos|>Once upon a time, there was a pale boy named Tim. Tim loved to
-
story 2.48time, there was a little boy named Tim. Tim was very
-
story 2.47`.<|assistant_end|><|bos|>Once upon a time, there was a little boy named Tim. Tim was troubled