Skip to content
mini-lab
Features of prelude-2

Feature #2194

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

Fires on
1.00% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 91%
  • python 3%
  • story 3%
  • chat 2%
  • worksheet 0%

Makes the model say

vowelsaeioureplaceAllitemsmainpositive defined counts

Makes it not say

def9 Tim saw8 The Tom loved

Strongest examples

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

  1. agent 2.68
    matches⏎items.py:⏎ Line 9: def count(items):⏎<|tool_end|><|assistant_start|>
  2. agent 2.63
    matches⏎core.py:⏎ Line 1: def count(items):⏎<|tool_end|><|assistant_start|>
  3. agent 2.6
    matches⏎lists.py:⏎ Line 8: def first(items):⏎<|tool_end|><|assistant_start|>
  4. agent 2.6
    matches⏎arith.py:⏎ Line 1: def square(x):⏎<|tool_end|><|assistant_start|>
  5. agent 2.58
    math_utils.py:⏎ Line 5: def half(x):⏎<|tool_end|><|assistant_start|>
  6. agent 2.58
    math_utils.py:⏎ Line 1: def previous(x):⏎<|tool_end|><|assistant_start|>
  7. agent 2.57
    math_utils.py:⏎ Line 5: def cube(x):⏎<|tool_end|><|assistant_start|>
  8. agent 2.55
    math_utils.py:⏎ Line 1: def twice(x):⏎<|tool_end|><|assistant_start|>
  9. agent 2.55
    number_utils.py:⏎ Line 1: def opposite(x):⏎<|tool_end|><|assistant_start|>
  10. agent 2.55
    tools.py:⏎ Line 1: def sub_one(x):⏎<|tool_end|><|assistant_start|>
  11. agent 2.5
    matches⏎utils.py:⏎ Line 1: def square(x):⏎<|tool_end|><|assistant_start|>
  12. agent 2.47
    number_utils.py:⏎ Line 1: def square(x):⏎<|tool_end|><|assistant_start|>