Skip to content
mini-lab
Log in Sign up
Features of prelude-2

Feature #1605

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

Fires on
0.23% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 72%
  • chat 21%
  • agent 4%
  • worksheet 2%
  • python 1%

Makes the model say

_<|arg|>ultipl addll import(of

Makes it not say

won fell05 woke1 from9

Strongest examples

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

  1. story 2.38
    quarrel anymore, and they were happy.<|bos|>Once upon a time, there was an
  2. story 2.3
    learned that sharing with others can make everyone happy.<|bos|>Once upon a time, there was a
  3. story 2.24
    from others. And they played happily ever after.<|bos|>Once upon a time, in a small
  4. story 2.22
    <|assistant_end|><|bos|>Once upon a time, there was a
  5. story 2.22
    friend again. And they lived happily ever after.<|bos|>Once upon a time, there was a
  6. story 2.21
    bunny tight and got her yummy muffin.<|bos|>Once upon a time, there was a
  7. story 2.21
    and fix the Python files in this project.<|assistant_end|><|bos|>Once upon a time, there was a
  8. story 2.2
    happy to be friends and to help each other.<|bos|>Once upon a time there was a young
  9. story 2.18
    Python files in this project.<|assistant_end|><|bos|>Once upon a time, there was a
  10. story 2.17
    <|think_end|>The answer is 13.<|assistant_end|><|bos|>Once upon a time, in a bright
  11. story 2.17
    old lady was happy.<|bos|>Once upon a time, there was a
  12. story 2.17
    to show everyone back home!<|bos|>Once upon a time, in a small