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

Feature #1415

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

Fires on
0.68% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 39%
  • agent 38%
  • story 16%
  • python 6%
  • worksheet 0%

Makes the model say

Hi upon laugh tell laughedOne answerer

Makes it not say

+:`.]`.)`.heses theirnew

Strongest examples

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

  1. chat 3.11
    ower?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  2. chat 2.95
    8?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  3. chat 2.9
    me.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  4. chat 2.89
    one?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  5. chat 2.89
    space.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  6. chat 2.84
    cook pasta<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  7. chat 2.83
    up?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  8. chat 2.83
    es?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  9. chat 2.82
    a cat<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  10. chat 2.81
    me.<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  11. chat 2.78
    code?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and
  12. chat 2.78
    a cat<|user_end|><|assistant_start|>Sorry, I'm a tiny model: I can only tell short stories and