Feature #913
Fires on “c” (not labeled yet: read the examples)
- Fires on
- 0.72% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 53%
- story 31%
- worksheet 7%
- chat 5%
- python 4%
Makes the model say
ationesion tall positiveoldStringingem
Makes it not say
now testsnums some bird`w spot
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 3.48went to the park to do some exercise. He loved
-
agent 3.22fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.17fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.15fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.14fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.14fetch, websearch, write, execute, lsp
-
agent 3.14fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.14fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.14fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.13fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.13fetch, websearch, write, execute<|system_end|><|system_start|>You
-
agent 3.12fetch, websearch, write, execute<|system_end|><|system_start|>You