Feature #2385
Fires on “Hello” (not labeled yet: read the examples)
- Fires on
- 0.19% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 36%
- python 23%
- story 22%
- chat 18%
- worksheet 0%
Makes the model say
Hello, projectrenreturn functions comeure
Makes it not say
In sc ofend important hard which what
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
python 5.33after.<|bos|>```python⏎def hello(name):⏎ return "Hello, " + name
-
chat 5.23.<|assistant_end|><|bos|><|system_start|>You are a helpful assistant.<|system_end|><|user_start|>Hello<|user_end|><|assistant_start|>Hello!
-
python 5.13bug": false}⏎⏎$ python3 script.py⏎Hello, Ada
-
agent 5.09Write a script greeting.py that prints "Hello, Zoe!".
-
python 5.03<|bos|># app.py⏎def hello(name):⏎ return "Hello, " + name
-
python 5.01.⏎⏎```python⏎def hello(name):⏎ return "Hello, " + name
-
python 4.95⏎# helpers.py⏎def welcome(name):⏎ return "Hello, " + name
-
story 4.92Bob hopped over to the cauliflower and said, "Hello, caul
-
python 4.9<|bos|># strings.py⏎def hello(name):⏎ return "Hello, " + name
-
python 4.89hello("Eva"))⏎⏎$ python3 main.py⏎Hello, Eva!
-
python 4.87<|bos|># tools.py⏎def hello(name):⏎ return "Hello, " + name
-
python 4.87"!"`.⏎⏎```python⏎def welcome(name):⏎ return "Hello, " + "