Feature #3109
Fires on “-” (not labeled yet: read the examples)
- Fires on
- 0.33% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 66%
- story 24%
- python 4%
- chat 4%
- worksheet 2%
Makes the model say
new):red)` fun instead` double
Makes it not say
we([listingb seekpython We
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.55<|assistant_end|><|tool_start|>Read file mathlib.py, lines 1-9⏎1:
-
agent 2.52<|assistant_end|><|tool_start|>Read file number_utils.py, lines 1-8⏎1:
-
agent 2.49<|assistant_end|><|tool_start|>Read file words.py, lines 1-6⏎1:
-
agent 2.48<|assistant_end|><|tool_start|>Read file greetings.py, lines 1-14⏎1
-
agent 2.46AI agent running in OpenCode, a coding
-
agent 2.46AI agent running in OpenCode, a coding
-
agent 2.46AI agent running in OpenCode, a coding
-
agent 2.46AI agent running in OpenCode, a coding
-
agent 2.38.py<|tool_call_end|><|assistant_end|><|tool_start|>Read file run.py, lines 1-14⏎1
-
agent 2.36.py<|tool_call_end|><|assistant_end|><|tool_start|>Read file hello.py, lines 1-4⏎1:
-
agent 2.35.py<|tool_call_end|><|assistant_end|><|tool_start|>Read file arith.py, lines 1-5⏎1:
-
agent 2.35.py<|tool_call_end|><|assistant_end|><|tool_start|>Read file greetings.py, lines 1-10⏎1