Skip to content
mini-lab
Features of prelude-2

Feature #3175

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

Fires on
1.17% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 42%
  • chat 35%
  • worksheet 14%
  • python 6%
  • story 3%

Makes the model say

to<|think_start|>( mustmain[::? me

Makes it not say

])) moral,owed."," next from

Strongest examples

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

  1. chat 2.61
    answer is 161.<|assistant_end|><|user_start|>Now add 12 to the result.<|user_end|>
  2. chat 2.51
    answer is 101.<|assistant_end|><|user_start|>And add 10 to that?<|user_end|><|assistant_start|>
  3. chat 2.51
    The answer is 11.<|assistant_end|><|user_start|>Now add 90 to the result.<|user_end|>
  4. chat 2.51
    The answer is 18.<|assistant_end|><|user_start|>Now add 41 to the result.<|user_end|>
  5. chat 2.46
    The answer is 11.<|assistant_end|><|user_start|>Now add 90 to the result.<|user_end|>
  6. chat 2.45
    <|assistant_start|>The answer is 9.<|assistant_end|><|user_start|>Now add 12 to the result.<|user_end|>
  7. chat 2.44
    The answer is 14.<|assistant_end|><|user_start|>Now add 63 to the result.<|user_end|>
  8. chat 2.43
    answer is 611.<|assistant_end|><|user_start|>Now add 43 to the result.<|user_end|>
  9. chat 2.42
    answer is 852.<|assistant_end|><|user_start|>And add 51 to that?<|user_end|><|assistant_start|>
  10. chat 2.42
    The answer is 96.<|assistant_end|><|user_start|>Now add 50 to the result.<|user_end|>
  11. chat 2.41
    answer is 121.<|assistant_end|><|user_start|>And add 51 to that?<|user_end|><|assistant_start|>
  12. chat 2.39
    answer is 110.<|assistant_end|><|user_start|>Now add 64 to the result.<|user_end|>