Feature #837
Fires on “ princess” (not labeled yet: read the examples)
- Fires on
- 0.52% of tokens
- Mostly in
- chat
- Layer
- after block 4
Where it fires
- chat 46%
- agent 33%
- story 15%
- python 4%
- worksheet 2%
Makes the model say
idea too % with for at nodded.
Makes it not say
list⏎ moralFixedrement Soonassertbox
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 2.49jumping from tree to tree.<|assistant_end|><|user_start|>I want a story about a princess!<|user_end|><|assistant_start|>Once
-
chat 2.48a helpful assistant.<|system_end|><|user_start|>Can you tell me a story about a princess?<|user_end|><|assistant_start|>Once
-
chat 2.38is 823.<|assistant_end|><|user_start|>I want a story about a princess!<|user_end|><|assistant_start|>Once
-
chat 2.35920.<|assistant_end|><|user_start|>Can you tell me a story about a princess?<|user_end|><|assistant_start|>Once
-
chat 2.3523.<|assistant_end|><|user_start|>Tell me a story about a dragon.<|user_end|><|assistant_start|>Once
-
chat 2.31Answer in one short sentence.<|system_end|><|user_start|>I want a story about a princess!<|user_end|><|assistant_start|>Once
-
chat 2.3<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>Tell me a story about a dragon.<|user_end|><|assistant_start|>Once
-
chat 2.29105.<|assistant_end|><|bos|><|user_start|>Write a short story about a princess.<|user_end|><|assistant_start|>Once
-
chat 2.29Answer in one short sentence.<|system_end|><|user_start|>I want a story about a princess!<|user_end|><|assistant_start|>Once
-
chat 2.25Answer in one short sentence.<|system_end|><|user_start|>I want a story about a princess!<|user_end|><|assistant_start|>Once
-
chat 2.25every answer with "Sure!".<|system_end|><|user_start|>Write a short story about a princess.<|user_end|><|assistant_start|>Sure
-
chat 2.24<|think_end|>509<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator.<|system_end|>