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

Feature #2512

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

Fires on
0.29% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 74%
  • agent 14%
  • story 6%
  • worksheet 4%
  • python 1%

Makes the model say

OneaOnce don gets're okayakes

Makes it not say

6 and anymore95,. The

Strongest examples

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

  1. chat 2.04
    asking me for a story!<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  2. chat 2.04
    short stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  3. chat 1.98
    or some numbers to add?<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  4. chat 1.96
    is 144.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  5. chat 1.92
    short stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  6. chat 1.9
    short stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  7. chat 1.89
    is 172.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  8. chat 1.88
    <|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  9. chat 1.87
    there was an old cat.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  10. chat 1.87
    happy and full. The end.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  11. chat 1.84
    story to tell his friends.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
  12. chat 1.83
    16250.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator