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

Feature #138

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

Fires on
0.51% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 65%
  • chat 20%
  • story 13%
  • python 1%
  • worksheet 0%

Makes the model say

car ball?"eret. gone min

Makes it not say

<|tool_call_start|> +geTogetherpathhelpers prelude ^^^^^^^^^^^^^^^^^^^^^^^^

Strongest examples

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

  1. chat 2.71
    orts scores.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  2. chat 2.63
    orts scores.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  3. chat 2.63
    sics to me.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  4. chat 2.63
    orts scores.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  5. chat 2.63
    o and Juliet?<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  6. chat 2.6
    sics to me.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  7. chat 2.59
    article for me.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  8. chat 2.54
    article for me.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  9. chat 2.52
    Berlin today?<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  10. chat 2.48
    o and Juliet?<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  11. chat 2.47
    ident of Canada?<|user_end|><|assistant_start|>Sorry, I don't know about that. I can
  12. chat 2.47
    sics to me.<|user_end|><|assistant_start|>Sorry, I don't know about that. I can