Feature #137
Fires on “ould” (not labeled yet: read the examples)
- Fires on
- 0.29% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 70%
- chat 21%
- story 9%
- python 0%
- worksheet 0%
Makes the model say
oreion_*.g counts questiononfig sentence
Makes it not say
." A saf happy saw bigAs
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 3.99in this project.<|assistant_end|><|bos|><|user_start|>Hey<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.75numbers.<|assistant_end|><|bos|><|user_start|>Good morning!<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.68Sure!".<|system_end|><|user_start|>Hi<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story
-
chat 3.6425.<|assistant_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.62friendly assistant.<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.62Sure!".<|system_end|><|user_start|>Hi<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story
-
chat 3.630.<|assistant_end|><|bos|><|user_start|>Hello!<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.55Sure!".<|system_end|><|user_start|>Hi<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story
-
chat 3.52!".<|system_end|><|user_start|>Hi!<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story
-
chat 3.46helpful assistant.<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.4<|bos|><|system_start|>tools: calculator<|system_end|><|user_start|>Hey<|user_end|><|assistant_start|>Hi there! Would you like a story
-
chat 3.38Sure!".<|system_end|><|user_start|>Hi<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story