Skip to content
mini-lab
Features of prelude-2

Feature #2778

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

Fires on
0.75% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 54%
  • story 30%
  • python 12%
  • chat 4%
  • worksheet 2%

Makes the model say

climbed == TodayONde wings climb from

Makes it not say

she ofe She you care also he

Strongest examples

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

  1. story 2.97
    John wanted to sail on the ocean, so they decided to build a boat out of
  2. story 2.94
    The answer is 135.<|assistant_end|><|bos|>Jen and Ben were twins who liked to
  3. story 2.91
    <|tool_end|><|assistant_start|>Created `greeting.py`.<|assistant_end|><|bos|>Sam and Ben are brothers. They
  4. story 2.83
    scratch on a laptop by mini-lab.<|assistant_end|><|bos|>Anna and Ben are friends. They like
  5. story 2.77
    a story, or some numbers to add?<|assistant_end|><|bos|>Sara and Tom are friends. They like
  6. story 2.74
    in `main.py`.<|assistant_end|><|bos|>Lily and Ben are playing in the park
  7. story 2.7
    , 4, 9])⏎9⏎<|bos|>Tom and Ann are friends. They like
  8. story 2.7
    can go outside and play with the pen. And later, we can write with a real pen
  9. story 2.63
    495 is 2603.<|bos|>Anna and Ben are friends. They like
  10. story 2.63
    <|tool_end|><|assistant_start|>The answer is 94.<|assistant_end|><|bos|>Lily and Ben are playing in the kitchen
  11. story 2.59
    8<|think_end|>The answer is 78.<|assistant_end|><|bos|>Tom and Sam are friends. They like
  12. story 2.59
    <|tool_end|><|assistant_start|>Created `say.py`.<|assistant_end|><|bos|>Anna and Ben are twins. They like