Skip to content
mini-lab
Features of prelude-2

Feature #367

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

Fires on
0.89% of tokens
Mostly in
python
Layer
after block 4

Where it fires

  • python 53%
  • agent 43%
  • chat 2%
  • story 2%
  • worksheet 1%

Makes the model say

"`. checks<|assistant_end|>inedreetoh ful

Makes it not say

Shes namedL " He They

Strongest examples

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

  1. python 3.16
    mark to s. For example, `exclaim("python")` returns `"python!"`.⏎⏎
  2. python 3.13
    mark to s. For example, `exclaim("banana")` returns `"banana!"`.⏎⏎
  3. python 3.13
    , Sam!"⏎assert excite("sun") == "sun!"⏎print("
  4. python 3.09
    ") == "pizza pizza"⏎assert excite("lemon") == "lemon!"⏎assert size
  5. python 3.07
    lower case. For example, `whisper("Banana")` returns `"banana"`.⏎⏎```
  6. python 3.07
    an exclamation mark. For example, `excite("banana")` returns `"banana!"`.⏎<|bos|>
  7. python 3.05
    "cloud"⏎assert whisper("BANANA") == "banana"⏎print("
  8. python 3.05
    banana") == "banana!"⏎assert exclaim("python") == "python!"⏎assert is
  9. python 3.03
    tree") == "tree!"⏎assert exclaim("sun") == "sun!"⏎print("
  10. python 3.0
    mark to s. For example, `exclaim("river")` returns `"river!"`.⏎⏎
  11. python 2.99
    mark to s. For example, `excite("hello")` returns `"hello!"`.⏎<|bos|>
  12. python 2.97
    ⏎```⏎⏎>>> whisper("BANANA")⏎"banana"⏎⏎`