Feature #2851
Fires on “ edit” (not labeled yet: read the examples)
- Fires on
- 0.53% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 81%
- chat 8%
- python 6%
- story 4%
- worksheet 1%
Makes the model say
longer tomorrow loveoodNow joides do
Makes it not say
with thought small =c equals the2
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 1.66on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.65on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.64on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.59on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.56on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.55on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.54<|arg|>oldString= return "Hello, " + name + "!"<|arg|>newString= return "
-
agent 1.53on a laptop. I can list, read, run, create, edit and fix the Python
-
agent 1.53on a laptop. I can list, read, run, create, edit and fix the Python
-
worksheet 1.53add 7 and
-
chat 1.53add numbers.<|assistant_end|><|bos|>
-
chat 1.53add 9 and