Feature #3753
Fires on โ.โ (not labeled yet: read the examples)
- Fires on
- 0.72% of tokens
- Mostly in
- chat
- Layer
- after block 4
Where it fires
- chat 77%
- agent 14%
- python 6%
- story 2%
- worksheet 1%
Makes the model say
<|user_end|>โOnce "ur<|assistant_start|> sun tight
Makes it not say
aloneOhHey` book found)`py
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 3.18step or with a calculator.<|assistant_end|><|bos|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>One day
-
chat 3.18short stories and add numbers.<|assistant_end|><|bos|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>Once upon
-
chat 3.18, or some numbers to add?<|assistant_end|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>One day
-
chat 3.09<|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>One day
-
chat 3.02't know about that. I can only tell short stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools
-
chat 3.01tell stories and add numbers.<|assistant_end|><|bos|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>Once upon
-
chat 2.97<|bos|><|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>One day
-
chat 2.97bye! Have a nice day.<|assistant_end|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>Once upon
-
chat 2.96't know about that. I can only tell short stories and add numbers.<|assistant_end|><|bos|>52
-
chat 2.96<|assistant_start|>The answer is 7.<|assistant_end|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>Once upon
-
chat 2.96<|system_start|>Answer in one short sentence.<|system_end|><|user_start|>Tell me a short story.<|user_end|><|assistant_start|>Once upon
-
chat 2.95by step or with a calculator.<|assistant_end|><|bos|><|user_start|>Tell me a story.<|user_end|><|assistant_start|>Once upon