Feature #122
Fires on “ run” (not labeled yet: read the examples)
- Fires on
- 3.22% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 48%
- chat 32%
- python 13%
- story 4%
- worksheet 3%
Makes the model say
in to for ideaitems k can //
Makes it not say
Thank moral<|assistant_end|>R thankful<|bos|> proud thanked
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 4.13in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Please run the tests and fix
-
agent 4.11AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>Please app.py cras
-
agent 3.87agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>Please run hello.py for me
-
agent 3.79running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Explain hello from words.py
-
agent 3.76OpenCode, a coding agent harness.<|system_end|><|user_start|>"Run run.py."<|user_end|>
-
agent 3.67AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>Please run the tests and fix
-
agent 3.65AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>Run run.py and fix
-
agent 3.57AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>Please main.py cras
-
agent 3.55and `final_letter(s)`.<|assistant_end|><|user_start|>Execute run.py.<|user_end|>
-
agent 3.55<|assistant_end|><|user_start|>What's the output of run.py?<|user_end|>
-
agent 3.51)` and `how_many(items)`.<|assistant_end|><|user_start|>Can you run hello.py?<|user_end|>
-
agent 3.5py` with `mean_list(nums)`.<|assistant_end|><|user_start|>Can you run script.py?