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

Feature #2022

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

Fires on
0.42% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 35%
  • chat 28%
  • worksheet 17%
  • python 12%
  • agent 9%

Makes the model say

gets python == want])() you+

Makes it not say

⏎ thankfuli ButJustwhat dark Everyone

Strongest examples

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

  1. worksheet 6.67
    ave a nice day.<|assistant_end|><|bos|>Tim has 86 shells. Tim gets 72
  2. worksheet 6.27
    330.<|assistant_end|><|bos|>Mia has 434 apples. Mia gets 24
  3. worksheet 6.03
    6490.<|bos|>Mia has 208 coins. Mia gets 97
  4. chat 5.96
    add to plus".<|assistant_end|><|bos|><|user_start|>Anna has 87 stars. Anna gets 87
  5. chat 5.92
    tests pass now.<|assistant_end|><|bos|><|user_start|>Lily has 71 coins. Lily gets 8 more
  6. chat 5.84
    use the calculator.<|system_end|><|user_start|>Mia has 917 stickers. Mia gets 0 more
  7. worksheet 5.8
    The tests pass now.<|assistant_end|><|bos|>Lucy has 67 shells. Lucy gets 47
  8. chat 5.75
    is 28.<|assistant_end|><|user_start|>Lily has 47 books. Lily gets 45
  9. chat 5.65
    Answer with the number only.<|system_end|><|user_start|>Anna has 6 stars. Anna gets 0 more
  10. worksheet 5.63
    hi.py`.<|assistant_end|><|bos|>Anna has 88 balls. Anna gets 4 more
  11. chat 5.63
    372.<|assistant_end|><|bos|><|user_start|>Lily has 71 coins. Lily gets 8 more
  12. chat 5.59
    the number only.<|system_end|><|user_start|>Lily has 371 apples. Lily gets 31