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

Feature #729

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

Fires on
0.28% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 43%
  • story 37%
  • agent 10%
  • worksheet 8%
  • python 2%

Makes the model say

Hello longerigh firstier out check else

Makes it not say

The. the they The EveryThey They

Strongest examples

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

  1. chat 2.57
    <|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  2. chat 2.57
    <|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  3. chat 2.53
    <|think_end|>The answer is 17.<|assistant_end|><|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  4. chat 2.33
    you today?<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  5. chat 2.32
    7 + 25 = 72<|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  6. chat 2.3
    can only tell short stories and add numbers.<|assistant_end|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  7. chat 2.26
    can only tell short stories and add numbers.<|assistant_end|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  8. chat 2.25
    <|user_start|>What are you good at?<|user_end|><|assistant_start|>
  9. chat 2.24
    cat together.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  10. chat 2.24
    something special that she had earned.<|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  11. chat 2.23
    The answer is 816.<|assistant_end|><|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>
  12. chat 2.21
    The answer is 630.<|assistant_end|><|bos|><|user_start|>What are you good at?<|user_end|><|assistant_start|>