Skip to content
mini-lab
Features of prelude-2

Feature #4034

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

Fires on
0.55% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 89%
  • chat 8%
  • story 3%
  • python 0%
  • worksheet 0%

Makes the model say

= They onlyRename are about uponpered

Makes it not say

my5$lii and little1

Strongest examples

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

  1. agent 3.81
    path=utils.py<|arg|>oldString= return a * b<|arg|>newString= return a
  2. agent 3.77
    .py<|arg|>oldString= return s.lower()<|arg|>newString= return s
  3. agent 3.77
    test_arith.py<|arg|>oldString=positive<|arg|>newString=is_positive
  4. agent 3.75
    = return s[-1]<|arg|>newString= return s
  5. agent 3.75
    =strings.py<|arg|>oldString= return s.upper()<|arg|>newString= return s
  6. agent 3.7
    =core.py<|arg|>oldString= return s.upper()<|arg|>newString= return s
  7. agent 3.68
    path=core.py<|arg|>oldString= return a - b<|arg|>newString= return a
  8. agent 3.64
    =test_calc.py<|arg|>oldString=absolute<|arg|>newString=magnitude<|arg|>
  9. agent 3.63
    path=ops.py<|arg|>oldString= return a * b<|arg|>newString= return a
  10. agent 3.63
    math_utils.py<|arg|>oldString= return a * b<|arg|>newString= return a
  11. agent 3.63
    path=calc.py<|arg|>oldString= return a * b<|arg|>newString= return a
  12. agent 3.62
    =helpers.py<|arg|>oldString= return s.upper()<|arg|>newString= return s