Feature #2085
Fires on “ you” (not labeled yet: read the examples)
- Fires on
- 0.59% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 72%
- chat 14%
- story 9%
- python 4%
- worksheet 1%
Makes the model say
seen{ done loud[:: picture functions defines
Makes it not say
."⏎Thank6?7est?"
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 6.05Hi!<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.9<|assistant_end|><|user_start|>Hi!<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.56mini-lab.<|assistant_end|><|user_start|>Hello<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.48this project.<|assistant_end|><|bos|><|user_start|>Hey<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.445.<|assistant_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
story 5.24ffball said, "My name is Poke. Would you like to be friends
-
story 5.19oey, I want to examine something with you. Would you help me?" Joe
-
chat 5.15assistant.<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.15assistant.<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.14with a calculator.<|assistant_end|><|user_start|>Hi<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.1and happy.<|assistant_end|><|user_start|>Hello!<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
chat 5.06<|system_start|>tools: calculator<|system_end|><|user_start|>Hey<|user_end|><|assistant_start|>Hi there! Would you like a story,