Skip to content
mini-lab
Features of prelude-2

Feature #2005

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

Fires on
0.56% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 39%
  • worksheet 30%
  • story 16%
  • agent 11%
  • python 3%

Makes the model say

does⏎ did could can liked ever do

Makes it not say

hedpositiveopB more longer heavy!"

Strongest examples

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

  1. chat 8.65
    has 8 toys. Sue gets 5 more toys. How many toys does Sue have now
  2. chat 8.64
    9 toys. Ben gets 551 more toys. How many toys does Ben have now
  3. chat 8.38
    10 toys. Ben gets 80 more toys. How many toys does Ben have now
  4. chat 8.36
    9 toys. Ben gets 551 more toys. How many toys does Ben have now
  5. worksheet 8.03
    764 toys. Lily gets 6 more toys. How many toys does Lily have now
  6. chat 8.01
    41 toys. Anna gets 37 more toys. How many toys does Anna have now
  7. chat 7.94
    91 coins. Sue gets 54 more coins. How many coins does Sue have now
  8. worksheet 7.93
    toys. Max gets 9723 more toys. How many toys does Max have now
  9. chat 7.89
    more flowers. How many flowers does Anna have now
  10. worksheet 7.88
    61 cookies. Max gets 1 more cookies. How many cookies does Max have now
  11. worksheet 7.87
    cookies. Lily gets 7589 more cookies. How many cookies does Lily have now
  12. worksheet 7.85
    has 6 toys. Max gets 2 more toys. How many toys does Max have now