Feature #1799
Fires on “One” (not labeled yet: read the examples)
- Fires on
- 0.55% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 72%
- chat 24%
- agent 2%
- python 2%
- worksheet 1%
Makes the model say
day afteritems, sunny] timeSo
Makes it not say
sebig biggerest anyonelownven
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 5.78<|bos|>One day, a little
-
story 5.5Lucy for making it look so pretty.<|bos|>One day, a little
-
story 5.49things and to be more careful with his toy weapon.<|bos|>One day, a cat
-
story 5.46when she went exploring.<|bos|>One day, a little
-
story 5.3829839 = 74763.<|bos|>One day, a little
-
chat 5.33<|assistant_end|><|bos|><|user_start|>I want a story about a cat!<|user_end|><|assistant_start|>One day, a big
-
story 5.32+ 73 = 82.<|bos|>One day, a little
-
story 5.31defined in `items.py`, line 1.<|assistant_end|><|bos|>One sunny day, a
-
story 5.27⏎So 4 + 89 = 93.<|bos|>One day, a big
-
chat 5.27.<|assistant_end|><|bos|><|user_start|>I want a story about a dog!<|user_end|><|assistant_start|>One day, Dog
-
story 5.26and made a new friend.<|bos|>One day, a bunny
-
story 5.252 + 475 = 477.<|bos|>One day, a little