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

Feature #3484

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

Fires on
1.33% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 66%
  • agent 18%
  • worksheet 9%
  • story 4%
  • python 2%

Makes the model say

you youritemsant x today H sum

Makes it not say

In table wordsock else same` around

Strongest examples

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

  1. chat 4.6
    model: I can tell short stories and add numbers.<|assistant_end|><|bos|><|user_start|>Could you tell me what
  2. chat 4.46
    utils.py` with `exclaim(s)`.<|assistant_end|><|bos|><|user_start|>I want a story!<|user_end|>
  3. chat 4.36
    for another story soon.<|assistant_end|><|bos|><|user_start|>I want a story about a
  4. chat 4.34
    . Now Lily has 64965 cookies.<|bos|><|user_start|>Could you tell me what
  5. chat 4.32
    0805 = 10807.<|bos|><|user_start|>Could you tell me what
  6. chat 4.25
    4507 is 64510.<|bos|><|user_start|>Could you tell me what
  7. chat 4.25
    that going to the store was a fun adventure.<|assistant_end|><|bos|><|user_start|>I want a story about the
  8. chat 4.15
    not tidy anymore. The big, mean monster had won.<|bos|><|user_start|>Could you tell me what
  9. chat 4.12
    , and they went home with big smiles on their faces.<|bos|><|user_start|>Could you tell me what
  10. chat 4.08
    (nums) / len(nums)⏎```⏎<|bos|><|user_start|>I want a story about a
  11. chat 4.05
    goodbye to his new friend and went home to get warm.<|bos|><|user_start|>Could you tell me what
  12. chat 4.02
    Sure! The answer is 417.<|assistant_end|><|bos|><|user_start|>I want a story!<|user_end|>