Feature #3855
Fires on “ a” (not labeled yet: read the examples)
- Fires on
- 0.58% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 72%
- chat 19%
- worksheet 4%
- agent 3%
- python 2%
Makes the model say
safe played happy loved fundef beautifulWow
Makes it not say
asked asks can togethere how do I
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 2.61<|bos|>One day, a little ball named B
-
story 2.55a story, or some numbers to add?<|assistant_end|><|bos|>One day, a strong boy named Tom
-
story 2.54spoiled apple was not so bad after all.<|bos|>One day, a small girl named Emma
-
story 2.51scary, but it brought them a nice surprise.<|bos|>One day, a little boy named Tim
-
story 2.51.<|assistant_end|><|bos|>One day, a young boy named Tim
-
story 2.5found the big, shiny balloon in the tree.<|bos|>One day, a little boy named Tim
-
story 2.5bravest animal in the whole jungle.<|bos|>One day, a grumpy bear named Ben
-
story 2.5when she went exploring.<|bos|>One day, a little boy named Tim
-
story 2.49ent and accepted Tom's help.<|bos|>One day, a little boy named Tim
-
story 2.49<|assistant_start|>The answer is 948.<|assistant_end|><|bos|>One day, a boy named Tim went
-
story 2.49helping him to prevent further danger.<|bos|>One day, a little girl named Amy
-
story 2.491656<|think_end|>1656<|assistant_end|><|bos|>One day, a wise old tree stood