Skip to content
mini-lab
Features of prelude-2

Feature #105

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

Fires on
1.44% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 47%
  • agent 42%
  • chat 8%
  • python 2%
  • worksheet 1%

Makes the model say

Finally until,Finally set to one man

Makes it not say

pathiverll noise adds /du"))

Strongest examples

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

  1. story 3.75
    and then seek me." Ben says, "OK. One, two, three..." Lily
  2. story 3.38
    very excited and started to count right away. He counted one, two, three, four,
  3. chat 3.15
    pumpkin out, but it was too tight. She tried again and again, but the pumpkin would
  4. story 3.1
    ly a little bit. On the count of three. One, two, three!"⏎Sara
  5. story 2.86
    to number their steps. They took turns counting one, two, three, and so
  6. story 2.81
    . They count the blocks as they add them. "One, two, three, four,
  7. story 2.76
    The bird did not give up. He tried again and again. Finally, the bird got the
  8. story 2.75
    the stairs. She counted them as she went. One, two, three, four,
  9. story 2.74
    tried to balance on it. He was persistent and tried again and again.⏎Finally
  10. chat 2.73
    get the cake. But he was too small. Tim tried and tried, but he could not
  11. story 2.71
    't look." Lily closed her eyes and counted. "One, two, three, four,
  12. story 2.71
    the chair, but it was too high. He tried again and again, but he could not