Feature #3636
Fires on “Sam” (not labeled yet: read the examples)
- Fires on
- 1.96% of tokens
- Mostly in
- worksheet
- Layer
- after block 4
Where it fires
- worksheet 56%
- chat 24%
- agent 14%
- story 4%
- python 1%
Makes the model say
has had listAdded sum adds prelude*
Makes it not say
the0 xs books a can it
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
worksheet 13.6146 + 0 = 46.⏎Lucy had 54
-
worksheet 13.412981 = 79424.⏎Mia has 8 books
-
worksheet 13.199 + 640 = 649.⏎Mia has 85
-
worksheet 13.0214 + 96 = 6610.⏎Anna had 75
-
worksheet 12.767684 = 13218.⏎Lucy had 33
-
worksheet 12.719 + 1 = 10.⏎Max has 6 toys
-
worksheet 12.6648698 = 49516.⏎Sam had 63
-
worksheet 12.5258681 = 58687.⏎Sam has 16
-
worksheet 12.366 plus 23 is 55269.⏎Lucy had 29
-
worksheet 12.28and 50, you get 5244.⏎Sam has 8 coins
-
worksheet 12.24465, you get 95100.⏎Mia has 82
-
worksheet 12.2+ 4312 = 4396.⏎Sam had 7 stickers