Feature #835
Fires on “ like” (not labeled yet: read the examples)
- Fires on
- 0.53% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 66%
- story 14%
- chat 12%
- python 7%
- worksheet 1%
Makes the model say
his himut treeassert worry look >
Makes it not say
⏎ not wish It legsIf thing wished
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 2.72.<|assistant_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.63-lab.<|assistant_end|><|user_start|>Hello<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.58project.<|assistant_end|><|bos|><|user_start|>Hey<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.48.<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.47calculator<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.47<|assistant_end|><|user_start|>Hi!<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.42a calculator.<|assistant_end|><|user_start|>Hi<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.42lab.<|assistant_end|><|user_start|>hi<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.42.<|assistant_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.4: calculator<|system_end|><|user_start|>Hi!<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.39.<|system_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story, or
-
chat 2.35<|assistant_end|><|bos|><|user_start|>Good morning!<|user_end|><|assistant_start|>Hi there! Would you like a story, or