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

Feature #2264

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

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

Where it fires

  • agent 40%
  • worksheet 25%
  • chat 19%
  • python 12%
  • story 4%

Makes the model say

red<|think_end|>myRenamedschema stories mini[-

Makes it not say

man now sawpython next sees Now heard

Strongest examples

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

  1. agent 1.55
    ⏎ if x < 0:⏎ return -x
  2. worksheet 1.53
    5=5, 580⏎0
  3. worksheet 1.53
    5=5, 5833⏎
  4. worksheet 1.4
    70 = 77.⏎
  5. worksheet 1.38
    1=2, 234⏎0
  6. worksheet 1.38
    0118 = 7386
  7. chat 1.37
    1=11, 113<|think_end|>The
  8. chat 1.37
    4=11, 1⏎12+
  9. story 1.37
    up to the stage. She met some other competitors but none of them seemed
  10. worksheet 1.36
    7=7, 77⏎0+
  11. worksheet 1.36
    28421 = 1110
  12. chat 1.33
    tall is the Eiffel Tower?<|user_end|><|assistant_start|>Sorry, I'm a tiny model: