Feature #509
Fires on “Do” (not labeled yet: read the examples)
- Fires on
- 0.76% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 47%
- chat 32%
- python 10%
- story 7%
- worksheet 5%
Makes the model say
blue checks<|system_end|>w Pythonularnd trained
Makes it not say
iceWe# feel0 do They import
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 2.79, 6, 3, 7, 7]) == 30⏎print
-
chat 2.54asking me for a story!<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
-
chat 2.48They had a great day.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
-
chat 2.42story to tell his friends.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
-
story 2.41a time there were two children, Joe and Ella. Joe had grown many ears of
-
python 2.39_list([6, 9, 7, 9])⏎3
-
chat 2.37is 172.<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
-
worksheet 2.3207 + 73 = 8880.⏎Hey, what's
-
worksheet 2.319 + 252 = 261.⏎Hey, what's
-
chat 2.31the hard work they had done!<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
-
chat 2.31ised that dreams can come true.<|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator
-
chat 2.309<|think_end|>509<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|system_start|>Do not use the calculator