Skip to content
mini-lab
Features of prelude-2

Feature #2572

Fires on “ Lily” (not labeled yet: read the examples)

Fires on
0.16% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 66%
  • agent 16%
  • chat 12%
  • python 3%
  • worksheet 3%

Makes the model say

are like⏎ opposite lovelettermax+

Makes it not say

entnew>>> there smell bought outway

Strongest examples

On text the dictionary never trained on. The underlined token is where it fires most.

  1. story 4.38
    37 = 91540<|bos|>Tom and Mia were best friends.
  2. story 4.36
    (nums)` returns every number of nums doubled.<|assistant_end|><|bos|>Tom and Lily went to the beach
  3. story 3.93
    507 = 92725.<|bos|>Lily and Tom are friends. They
  4. story 3.91
    : I can tell short stories and add numbers.<|assistant_end|><|bos|>Tom and Mia were playing in the
  5. story 3.82
    defines 1 function: `absolute(x)`.<|assistant_end|><|bos|>Tom and Lily were playing in the
  6. story 3.76
    textlib.py` and `ops.py`.<|assistant_end|><|bos|>Tom and Lily were playing in their
  7. story 3.74
    in `main.py`.<|assistant_end|><|bos|>Lily and Ben are playing in the
  8. story 3.69
    know how to tell stories and add numbers.<|assistant_end|><|bos|>Anna and Lily were best friends.
  9. story 3.65
    6, 607<|think_end|>607<|assistant_end|><|bos|>Ben and Lily liked to fish with
  10. story 3.61
    items)`. The tests pass now.<|assistant_end|><|bos|>Lily and Ben are playing with blocks
  11. story 3.59
    I can only tell short stories and add numbers.<|assistant_end|><|bos|>Lily and Ben were playing knight
  12. story 3.59
    utils.py` with `increment(x)`.<|assistant_end|><|bos|>Ben and Lily were playing in the