Feature #3004
Fires on “ a” (not labeled yet: read the examples)
- Fires on
- 0.22% of tokens
- Mostly in
- story
- Layer
- after block 4
Where it fires
- story 44%
- chat 33%
- agent 20%
- python 2%
- worksheet 2%
Makes the model say
both the t first long big door7
Makes it not say
calculatorselferheloudselveshellorite
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 2.56go there and play.
-
chat 2.56go home.⏎Be
-
chat 2.47friends. And they all lived happily ever after.<|bos|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>
-
chat 2.41together.<|bos|><|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>
-
chat 2.41, Lucy was happy to go home and rest.<|bos|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>
-
chat 2.41like a story, or some numbers to add?<|assistant_end|><|user_start|>Tell me a short story.<|user_end|>
-
chat 2.4can learn and have fun at the same time.<|bos|><|user_start|>Tell me a fact about space.
-
chat 2.4friends.<|bos|><|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a story about a cat
-
chat 2.32.<|assistant_end|><|bos|><|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>
-
chat 2.31together.<|bos|><|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>
-
chat 2.29145.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>Tell me a fact about space.
-
chat 2.28again.<|bos|><|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>