Feature #177
Fires on “7” (not labeled yet: read the examples)
- Fires on
- 1.60% of tokens
- Mostly in
- worksheet
- Layer
- after block 4
Where it fires
- worksheet 56%
- chat 31%
- python 6%
- agent 4%
- story 2%
Makes the model say
`+728_! anymore
Makes it not say
trainedpowergeill examplenve smallest
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
worksheet 6.94+ 5 = 9315.⏎5 plus 89 is 9
-
worksheet 6.922 = 112226.⏎4 plus 33 is 3
-
worksheet 6.867 + 13 = 20.⏎9 plus 2 equals 11
-
worksheet 6.853 = 144022.⏎3 plus 9976 is
-
worksheet 6.85you get 12546.⏎5 plus 783 equals
-
worksheet 6.84744 = 1592.⏎5 plus 512 is
-
worksheet 6.76+ 6 = 8877.⏎4 plus 6 is 10
-
worksheet 6.761 + 62 = 63.⏎2 plus 76146
-
worksheet 6.7176 = 53199.⏎4 plus 75 equals 7
-
worksheet 6.774 = 35269.⏎9 plus 5 is 14
-
worksheet 6.6227 = 80034.⏎7 plus 4507 is
-
worksheet 6.684 = 3676.⏎1 plus 96223