Skip to content
mini-lab
Features of prelude-2

Feature #3557

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

Fires on
0.67% of tokens
Mostly in
story
Layer
after block 4

Where it fires

  • story 64%
  • chat 28%
  • worksheet 6%
  • agent 1%
  • python 1%

Makes the model say

xtop_ ideaves yours setAt

Makes it not say

2nd count see$ pointble number

Strongest examples

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

  1. story 5.21
    )` returns x squared.<|assistant_end|><|bos|>One day, a grumpy cat named Tom went to the
  2. story 5.18
    files in this project.<|assistant_end|><|bos|>One day, a mighty big cat named Tom went for a
  3. story 5.04
    = 93.<|bos|>One day, a big, honest dog named Max went for a
  4. story 5.01
    84<|think_end|>584<|assistant_end|><|bos|>One day, a big cat named Tom wanted to go
  5. chat 4.95
    <|user_start|>Tell me a story.<|user_end|><|assistant_start|>One day, a grumpy cat named Mark had a t
  6. story 4.87
    is 105.<|assistant_end|><|bos|>One day, a cool cat named Sam went for a
  7. story 4.81
    days in the jungle.<|bos|>One day, a big, thick cat named Tom went for a
  8. story 4.73
    careful with his toy weapon.<|bos|>One day, a cat named Bow was walking
  9. story 4.72
    5 = 402.<|bos|>One day, a fast cat named Speedy
  10. story 4.71
    man. They are happy again.<|bos|>One day, a nosy cat named Tom saw a ho
  11. story 4.68
    eat. What a happy ending!<|bos|>One day, a cool cat named Tom went to a
  12. story 4.65
    happy and had lots of fun.<|assistant_end|><|bos|>One day, a cat named Jazz went for