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

Feature #809

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

Fires on
0.52% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 32%
  • worksheet 29%
  • story 18%
  • chat 16%
  • python 6%

Makes the model say

has'sprint haduallyrementnyeven

Makes it not say

_ak Fin faces happy Ben yescks

Strongest examples

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

  1. worksheet 5.62
    14 + 96 = 6610.⏎Anna had 75
  2. worksheet 5.42
    9 + 640 = 649.⏎Mia has 85
  3. worksheet 5.31
    2981 = 79424.⏎Mia has 8 books
  4. worksheet 5.19
    48698 = 49516.⏎Sam had 63
  5. worksheet 5.19
    19178 = 20700.⏎Lucy had 61
  6. worksheet 5.18
    4 + 12 = 80126.⏎Ben has 32
  7. worksheet 5.18
    63231 is 66534.⏎Anna has 81
  8. worksheet 5.14
    46 + 0 = 46.⏎Lucy had 54
  9. worksheet 5.1
    58681 = 58687.⏎Sam has 16
  10. worksheet 5.09
    29 + 5 = 75834.⏎Mia had 35
  11. worksheet 5.09
    6958 = 18888.⏎Anna has 61
  12. worksheet 5.06
    and 50, you get 5244.⏎Sam has 8 coins