Feature #53
Fires on “ welcome” (not labeled yet: read the examples)
- Fires on
- 0.27% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 66%
- story 13%
- python 10%
- chat 10%
- worksheet 1%
Makes the model say
course (m$ worryne twinstotal
Makes it not say
what dayent importing has said had
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 2.87<|system_end|><|user_start|>Thank you<|user_end|><|assistant_start|>Sure! You're welcome!<|assistant_end|><|bos|><|system_start|>
-
chat 2.82<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>Thank you<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|user_start|>I
-
chat 2.66<|assistant_end|><|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>Thank you<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|user_start|>what
-
chat 2.6kitchen table.<|assistant_end|><|user_start|>Thank you very much!<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|bos|>```
-
chat 2.56py⏎all tests passed⏎<|bos|><|user_start|>Thank you<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|user_start|>Who
-
chat 2.56mini-lab.<|assistant_end|><|bos|><|user_start|>thanks<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|user_start|>t
-
chat 2.51133.<|assistant_end|><|user_start|>thanks<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|bos|><|system_start|>
-
chat 2.5102.<|assistant_end|><|bos|><|user_start|>thanks<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|user_start|>What
-
chat 2.49989.<|assistant_end|><|user_start|>Thanks!<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|bos|>Once
-
chat 2.470.<|assistant_end|><|user_start|>Great, thanks!<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|user_start|>3
-
chat 2.47<|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>thanks<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|bos|>Sam
-
agent 2.46(a, b)`.<|assistant_end|><|user_start|>Thank you!<|user_end|><|assistant_start|>You're welcome!<|assistant_end|><|bos|><|user_start|>