Skip to content
mini-lab
Features of prelude-2

Feature #2620

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

Fires on
0.83% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 37%
  • story 29%
  • chat 25%
  • python 7%
  • worksheet 2%

Makes the model say

problem piece keepHiner othersood

Makes it not say

"!"ppedreturn hide maybe Tryhed hiding

Strongest examples

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

  1. agent 3.64
    , in <module>⏎ assert lower_case("PYTHON") == "
  2. chat 3.4
    pass now.<|assistant_end|><|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>What
  3. chat 3.24
    ready for another day of fun together.<|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>what
  4. chat 3.17
    2 is 1534.<|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>what
  5. python 3.12
    s.lower()⏎```⏎⏎>>> whisper("PYTHON")⏎"
  6. chat 3.1
    she learned something new with her mom.<|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>Could
  7. chat 3.08
    together, and they became good friends.<|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>What
  8. chat 3.03
    <|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>Sue
  9. chat 3.03
    <|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>Find
  10. chat 3.01
    and not shaken the tree.<|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>What
  11. chat 3.01
    ⏎```<|assistant_end|><|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>Calculate
  12. chat 2.99
    so much fun with their new toy.<|bos|><|system_start|>Answer with the number only.<|system_end|><|user_start|>I