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

Feature #2482

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

Fires on
0.40% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 52%
  • chat 30%
  • agent 14%
  • worksheet 3%
  • python 2%

Makes the model say

you yourOnce they there so here two

Makes it not say

an f ideaestI1 ballto

Strongest examples

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

  1. story 9.38
    walks slowly to the bird and says, "Hello, bird. What are you? You are
  2. chat 9.38
    .⏎The princess smiled and said "Hello, little puppy. What are you doing up here
  3. chat 9.36
    .⏎The princess smiled and said "Hello, little puppy. What are you doing up here
  4. story 9.31
    the slide. He says, "Hi, I'm Ben. What are you singing?"⏎
  5. story 9.3
    him. The bird said, "Hello, Mr. Bear! What are you doing on that
  6. story 9.19
    a crab. The crab said, "Hello, seal! What are you doing?" The
  7. story 9.07
    Chirpy. Chirpy said, "Hello, Quick! What are you doing up here
  8. story 8.89
    saw her playing with her kitchen.⏎"Hi, Amy! What are you doing?" asked
  9. story 8.88
    a bird. The bird said, "Hi, Will! Why are you so excite?"
  10. story 8.84
    dad walked up to him.⏎"Hey there son. What are you doing?"⏎
  11. story 8.81
    so he went up to the guard and said, "Hi, why are you here?"⏎
  12. chat 8.71
    trained from scratch on a laptop by mini-lab.<|assistant_end|><|user_start|>Who are you?<|user_end|><|assistant_start|>