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

Feature #2387

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

Fires on
0.57% of tokens
Mostly in
python
Layer
after block 4

Where it fires

  • python 51%
  • story 28%
  • agent 10%
  • worksheet 5%
  • chat 5%

Makes the model say

includes listpyMEtxtg forgultipl

Makes it not say

statsigage it you eyes stats strings

Strongest examples

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

  1. agent 2.06
    ent, webf
  2. agent 2.06
    ent, webf
  3. agent 2.06
    ent, webf
  4. python 1.86
    tree") == "t
  5. chat 1.76
    ice to meet you,
  6. story 1.75
    ing longer and his body was becoming ligh
  7. python 1.7
    >>> smaller(3, 3)⏎3⏎⏎`double(x)`
  8. python 1.69
    returns `[10, 14, 2]`.⏎⏎`sum_two(
  9. python 1.69
    case("apple")` returns `"APPLE"`.⏎⏎`join_words(
  10. python 1.68
    8, 0, 7])` returns `7`.⏎⏎`count_items(
  11. python 1.68
    ```⏎<|bos|>Bug: `mean_list`
  12. python 1.68
    6, 7, 7])` returns `[6]`.⏎⏎`total(nums)`