Feature #1968
Fires on “ day” (not labeled yet: read the examples)
- Fires on
- 0.06% of tokens
- Mostly in
- chat
- Layer
- after block 4
Where it fires
- chat 67%
- story 27%
- agent 5%
- python 1%
- worksheet 0%
Makes the model say
ran up top goodbyetCc So
Makes it not say
testsreturn them") +even line say
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 2.56?<|assistant_end|><|user_start|>I want a story about the park!<|user_end|><|assistant_start|>One day, a boy named
-
chat 2.48.<|assistant_end|><|user_start|>Tell me a story about a tree.<|user_end|><|assistant_start|>One day, a large dog
-
chat 2.46<|user_start|>Can you tell me a story about the sun?<|user_end|><|assistant_start|>One day, a little boy
-
chat 2.44to add?<|assistant_end|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>One day, a girl named
-
chat 2.4.<|assistant_end|><|user_start|>I want a story about a dog!<|user_end|><|assistant_start|>One day, a good dog
-
chat 2.4.<|system_end|><|user_start|>I want a story about the sea!<|user_end|><|assistant_start|>One day, a funny fish
-
chat 2.38help!<|assistant_end|><|user_start|>Can you tell me a story?<|user_end|><|assistant_start|>One day, a boy named
-
chat 2.37.<|system_end|><|user_start|>I want a story about a bunny!<|user_end|><|assistant_start|>One day, a large bunny
-
chat 2.37one short sentence.<|system_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>One day, a grumpy cat
-
chat 2.36<|user_start|>Can you tell me a story about a fish?<|user_end|><|assistant_start|>One day, a little fish
-
chat 2.35!<|assistant_end|><|user_start|>I want a story about the park!<|user_end|><|assistant_start|>One day, a little boy
-
chat 2.35?<|assistant_end|><|user_start|>I want a story about a bird!<|user_end|><|assistant_start|>One day, a little bird