Skip to content
mini-lab
Features of prelude-2

Feature #80

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

Fires on
1.27% of tokens
Mostly in
agent
Layer
after block 4

Where it fires

  • agent 54%
  • python 37%
  • story 6%
  • chat 2%
  • worksheet 1%

Makes the model say

Thankn named shiny * BlesR

Makes it not say

One! when coming passed come to run

Strongest examples

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

  1. story 3.38
    surprised because it hadn't taken
  2. story 2.67
    sad that Spot had been bad, but he was relieved that he hadn't gotten sick
  3. python 2.63
    <|bos|># main.py⏎def evens(nums):⏎ return [n for n in nums
  4. python 2.62
    main.py⏎def only_even(nums):⏎ return [n for n in nums
  5. python 2.62
    main.py⏎def keep_even(nums):⏎ return [n for n in nums
  6. story 2.58
    Mummy, but he still felt a little frustrated that the lemon hadn't fit in the
  7. story 2.54
    so sad. It thought it had made a friend. But it hadn't. The child
  8. agent 2.5
    py<|arg|>content=def only_even(nums):⏎ return [n for n in nums
  9. python 2.48
    .⏎⏎```python⏎def evens(nums):⏎ return [n for n in nums
  10. python 2.48
    ⏎⏎```python⏎def evens(nums):⏎ return [n for n in nums
  11. python 2.47
    items.py⏎def only_even(nums):⏎ return [n for n in nums
  12. python 2.46
    .⏎⏎```python⏎def evens(nums):⏎ return [n for n in nums