Feature #1683
Fires on “ watch” (not labeled yet: read the examples)
- Fires on
- 0.38% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 76%
- story 20%
- chat 2%
- python 1%
- worksheet 1%
Makes the model say
files list you people<|assistant_end|><|tool_start|>s double
Makes it not say
Don upon# won8ustisher
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 2.49girl named Emma wanted to enter the big room where her mom was watching a film. She
-
story 2.46One night, Jane decided to go to the stadium to watch the game. As
-
agent 2.46most recent call last):⏎ File "main.py", line 4, in <module
-
story 2.41Every day, he would sit in front of the TV and watch his favorite cartoon about a funny
-
agent 2.37call last):⏎ File "test_arith.py", line 5, in <module
-
agent 2.36call last):⏎ File "test_calc.py", line 3, in <module
-
story 2.33was a gloomy room. The room had old furniture that no one liked. One
-
agent 2.33call last):⏎ File "test_listlib.py", line 4, in <module
-
story 2.32. Her mom said, "We can bake the cake after we watch a movie
-
agent 2.31call last):⏎ File "test_utils.py", line 3, in <module
-
agent 2.3):⏎ File "test_textlib.py", line 6, in <module
-
story 2.27that he had come to the beach to bury the violin so that no one else could keep