Feature #2728
Fires on “1” (not labeled yet: read the examples)
- Fires on
- 0.30% of tokens
- Mostly in
- worksheet
- Layer
- after block 4
Where it fires
- worksheet 39%
- agent 31%
- chat 20%
- python 8%
- story 2%
Makes the model say
Fixboarditemcommand= typo:`nums
Makes it not say
thisOnce they back new returningsum
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
chat 2.8Hi!<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
worksheet 2.6766⏎00+36: 0+6+1=7,
-
worksheet 2.60+1: 0+1+1=2,
-
worksheet 2.57076⏎0+6: 0+6+1=7,
-
worksheet 2.5658⏎00+66: 0+6+1=7,
-
chat 2.535.<|assistant_end|><|user_start|>hey there<|user_end|><|assistant_start|>Hi there! Would you like a story,
-
worksheet 2.536: 0+6+1=7,
-
worksheet 2.5: 0+6+1=7,
-
chat 2.49, 63⏎0+6: 0+6+1=7,
-
chat 2.49!".<|system_end|><|user_start|>Hello<|user_end|><|assistant_start|>Sure! Hi there! Would you like a story,
-
worksheet 2.480, 0⏎6+0: 6+0+1=7,
-
chat 2.44, 67⏎6+0: 6+0+1=7,