Skip to content
mini-lab
Features of prelude-2

Feature #1453

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

Fires on
0.62% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 62%
  • agent 18%
  • chat 10%
  • python 6%
  • worksheet 3%

Makes the model say

<|user_start|> run list(" python Try moral+

Makes it not say

??".s⏎⏎x are!"

Strongest examples

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

  1. story 2.58
    was worth it to keep her toy safe. And from that day on, Lily always made
  2. story 2.51
    peedy. Speedy loved to go fast. He would speed down the street
  3. story 2.48
    was happy she made new friends in the network. From that day on, the park became
  4. story 2.47
    is good to show people we are happy for them. From that day on, Tom and Sue
  5. story 2.46
    He thanked his dad and gave him a big hug. From that day on, Sam was never
  6. story 2.43
    be home. He thanked Elly for her help. From that day on, Elly and
  7. story 2.43
    if you help others, you will feel good too. From that day on, all the animals
  8. story 2.42
    He thanked Squeaky and they became good friends. From that day on, Bobo was careful
  9. story 2.4
    could help his mom and dad talk to each other. From that day on, Tim loved both
  10. story 2.4
    very happy that her mom was proud of her.⏎From that day on, Lucy seemed to
  11. story 2.38
    big hug and put her back in the toy box. From that day on, Lily made sure
  12. story 2.38
    little dog felt better and was happy to be home. From that day on, the big dog