Feature #3922
Fires on “<|assistant_end|>” (not labeled yet: read the examples)
- Fires on
- 2.14% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 81%
- chat 7%
- worksheet 5%
- python 4%
- story 2%
Makes the model say
Hioo're is ItHello<|tool_call_end|>case
Makes it not say
instead both projectemember more Many the moral
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.34hello.<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def hello<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.26do?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def whisper<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.26smallest.<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def smallest<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.23hello."<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def hello<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.22larger.<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def larger<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.21whisper?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def whisper<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.15tail?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def tail<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.15return?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def whisper<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.14return?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def halve<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.13<|assistant_start|><|tool_call_start|>write<|arg|>path=tools.py<|arg|>content=def is_divisible(a
-
agent 2.13exclaim?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def exclaim<|tool_call_end|><|assistant_end|><|tool_start|>Found 1
-
agent 2.13do?<|user_end|><|assistant_start|><|tool_call_start|>grep<|arg|>pattern=def squared<|tool_call_end|><|assistant_end|><|tool_start|>Found 1