Skip to content
mini-lab
Features of prelude-2

Feature #884

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

Fires on
1.16% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 89%
  • story 5%
  • chat 3%
  • python 2%
  • worksheet 0%

Makes the model say

x<|tool_call_end|>if examplecaseassert<|think_end|>()

Makes it not say

i more( they kind other again average

Strongest examples

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

  1. agent 2.31
    pretty river.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  2. agent 2.31
    a walk.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  3. agent 2.28
    proud of her balance.<|bos|><|system_start|>tools: edit, glob, grep, question,
  4. agent 2.26
    the yard.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  5. agent 2.25
    greetings.py<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  6. agent 2.24
    big boat.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  7. agent 2.24
    a sailor.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  8. agent 2.24
    .py with bigger<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question,
  9. agent 2.23
    and add numbers.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question,
  10. agent 2.23
    and add numbers.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question,
  11. agent 2.23
    named Tim.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read
  12. agent 2.21
    a walk.<|assistant_end|><|bos|><|system_start|>tools: edit, glob, grep, question, read