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

Feature #1212

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

Fires on
0.69% of tokens
Mostly in
chat
Layer
after block 4

Where it fires

  • chat 44%
  • agent 25%
  • worksheet 15%
  • story 8%
  • python 8%

Makes the model say

): tests + <ERfdivisibleue

Makes it not say

was way got opened went climb found played

Strongest examples

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

  1. chat 3.65
    has 8 toys. Sue gets 5 more toys. How many toys does Sue have now
  2. chat 3.59
    10 toys. Ben gets 80 more toys. How many toys does Ben have now
  3. chat 3.58
    19 toys. Sam gets 38 more toys. How many toys does Sam have now
  4. chat 3.54
    41 toys. Anna gets 37 more toys. How many toys does Anna have now
  5. chat 3.38
    9 toys. Ben gets 551 more toys. How many toys does Ben have now
  6. chat 3.34
    41 toys. Anna gets 37 more toys. How many toys does Anna have now
  7. chat 3.32
    has 3 cookies. Sue gets 4 more cookies. How many cookies does Sue have now
  8. chat 3.3
    31 toys. Tom gets 4 more toys. How many toys does Tom have now
  9. worksheet 3.27
    167 toys. Sam gets 4 more toys. How many toys does Sam have now
  10. chat 3.24
    19 toys. Sam gets 38 more toys. How many toys does Sam have now
  11. chat 3.24
    has 4 toys. Tim gets 6 more toys. How many toys does Tim have now
  12. chat 3.24
    19 cookies. Tim gets 76 more cookies. How many cookies does Tim have now