Skip to content
mini-lab
Features of prelude-2

Feature #2454

Fires on “9” (not labeled yet: read the examples)

Fires on
1.60% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 55%
  • worksheet 30%
  • chat 13%
  • story 1%
  • python 0%

Makes the model say

6ident9 +7ch5 fall

Makes it not say

(- She Lily"`.et?" name

Strongest examples

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

  1. agent 3.29
    : ⏎9: def factorial(
  2. agent 3.04
    return items[-1]⏎7: ⏎8: ⏎9: NUMBERS = [
  3. agent 2.99
    : ⏎5: LIMIT =
  4. agent 2.98
    return s.upper()⏎7: ⏎8: ⏎9: def count_
  5. agent 2.96
    2 == 1]⏎7: ⏎8: ⏎9: def count(
  6. agent 2.94
    6: return k⏎7: ⏎8: ⏎9: def quiet(
  7. agent 2.94
    return items[1]⏎7: ⏎8: ⏎9: def has_
  8. agent 2.92
    8: ⏎9: def raise_
  9. agent 2.92
    8: ⏎9: WORD = "
  10. agent 2.91
    return x in items⏎7: ⏎8: ⏎9: def count(
  11. agent 2.91
    6: return k⏎7: ⏎8: ⏎9: def first_
  12. agent 2.91
    return s[0]⏎7: ⏎8: ⏎9: def excite(