Feature #2505
Fires on “2” (not labeled yet: read the examples)
- Fires on
- 1.48% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 76%
- python 16%
- story 3%
- worksheet 3%
- chat 2%
Makes the model say
42,0 play56ice
Makes it not say
]) len")_Setlib python exclamation
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 5.54line 3, in <module>⏎ assert max_item([0, 6,
-
agent 5.458, in <module>⏎ print(adda_ll([7, 9,
-
agent 5.379, 3]))<|arg|>newString=print(add_all([7, 9,
-
agent 5.335, in <module>⏎ print(taail([1, 7,
-
agent 5.31, in <module>⏎ print(ha_sitem([4, 7,
-
agent 5.25<|arg|>newString=print(max_item([5, 2,
-
agent 5.225, in <module>⏎ print(max_itm([2, 3,
-
agent 5.223, in <module>⏎ print(lagrest([2, 2,
-
agent 5.16<|arg|>path=main.py<|arg|>oldString=print(fist([2, 0,
-
agent 5.15line 3, in <module>⏎ assert min_item([2, 6,
-
agent 5.13<|arg|>path=script.py<|arg|>oldString=print(smallst([5, 8,
-
agent 5.12line 4, in <module>⏎ assert min_item([1, 5,