Feature #2814
Fires on “"” (not labeled yet: read the examples)
- Fires on
- 0.23% of tokens
- Mostly in
- agent
- Layer
- after block 4
Where it fires
- agent 76%
- story 11%
- chat 7%
- worksheet 4%
- python 1%
Makes the model say
ed4de wantasticfulodd
Makes it not say
le()`. element item lists(-name tests
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
agent 2.77agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"How does smallest work
-
agent 2.77agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"test_words.
-
agent 2.77agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Open mathlib.
-
agent 2.77agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Tell me about yourself
-
agent 2.77agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"use 8
-
agent 2.76an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Fix the bug caught
-
agent 2.76an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"In number_utils
-
agent 2.75an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Which Python files are
-
agent 2.75an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Run hello.py
-
agent 2.75an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Please add a function
-
agent 2.75an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"test_listlib.
-
agent 2.74an AI agent running in OpenCode, a coding agent harness.<|system_end|><|user_start|>"Why does hello.