Skip to content
mini-lab
Features of prelude-2

Feature #959

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

Fires on
0.44% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 66%
  • worksheet 13%
  • chat 9%
  • agent 8%
  • python 5%

Makes the model say

$ upon sorry toldnitude fixvowels if

Makes it not say

add thetitemsstdoto

Strongest examples

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

  1. story 2.07
    . The cat could not help him. Tim's day had a bad ending.<|bos|><|system_start|>tools
  2. story 2.05
    bird. I love animals. I want to be a veterinarian."⏎Her
  3. story 2.04
    <|bos|>Jenny loved animals. She wanted to be a veterinarian when she grew
  4. story 1.99
    not fun anymore. They were all sad because the party had a bad ending.<|bos|><|system_start|>tools
  5. story 1.91
    't be afraid. I'm Anna. I'm a veterinarian. I want
  6. story 1.75
    of the rain. They were sad that their fun day had a bad ending.<|bos|><|system_start|>tools
  7. story 1.74
    and a strange smell that Smiley had not encountered before.
  8. story 1.7
    big monster came out and scared them more. The show had a bad ending, and Lucy and
  9. story 1.59
    came out. They saw what happened. They called the veterinarian. The v
  10. story 1.59
    . They did not like the cat. The cat was mean and
  11. story 1.58
    "Well George, it's a bit complicated to explain.
  12. story 1.57
    to anyone else. We'll take him to the real veterinarian tomorrow and see