Skip to content
mini-lab
Features of prelude-2

Feature #2657

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

Fires on
0.34% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 44%
  • story 41%
  • worksheet 8%
  • chat 7%
  • python 1%

Makes the model say

agreed import!"ious ch excited= blue

Makes it not say

fixake Pythonopxt])) back7

Strongest examples

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

  1. story 2.69
    One day, Daddy gave Daisy a pedal bike. Daisy was so happy and couldn't
  2. story 2.65
    While he was walking, Tim met a friendly dog. The dog wagged its tail and barked.
  3. story 2.59
    Joey and his mummy were at the beach. Joey was so excited to build sand
  4. story 2.55
    , a big, friendly dog came up to him. The dog wagged its tail and barked at
  5. story 2.53
    ⏎One day Jack found a print of a beautiful rainbow. He was so lucky! He wanted
  6. story 2.41
    day, the cat found a bag of rice. The cat was so excited! It started
  7. story 2.41
    the package and find out what's inside!"⏎Fluffy was so excited it raced up
  8. story 2.37
    ized to him and invited him to play with them. Spot wagged his tail and licked their
  9. story 2.36
    a cheap cell phone for Mary to play with. Mary was so excited to play with
  10. story 2.35
    . You are cute. Can I pet you?"⏎The dog wagged his tail. He licked
  11. story 2.34
    Mary, we don't have any beef." Mary was so sad and said,
  12. story 2.34
    present, it was a pastel colour! Gemma was so excited that she jumped