Feature #2375
Fires on โ<|bos|>โ (not labeled yet: read the examples)
- Fires on
- 0.74% of tokens
- Mostly in
- chat
- Layer
- after block 4
Where it fires
- chat 30%
- story 25%
- agent 25%
- python 10%
- worksheet 10%
Makes the model say
fromnumber number[- +<|think_end|> of],
Makes it not say
moralt new gave anymore ( that taught
Strongest examples
On text the dictionary never trained on. The underlined token is where it fires most.
-
story 20.29they had a fun day playing with the treasure in the park.<|assistant_end|><|bos|>Once upon a time
-
agent 20.2that. I only know how to tell stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: edit
-
agent 20.1that. I only know how to tell stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: edit
-
agent 20.08that. I only know how to tell stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: edit
-
chat 20.04that. I only know how to tell stories and add numbers.<|assistant_end|><|bos|><|user_start|>Who made you
-
chat 20.02about that. I can only tell short stories and add numbers.<|assistant_end|><|bos|><|user_start|>Can you write
-
agent 20.02<|tool_call_end|><|assistant_end|><|tool_start|>all tests passedโ<|tool_end|><|assistant_start|>All tests pass.<|assistant_end|><|bos|><|system_start|>tools: edit
-
chat 20.02instead of `if a > b:`. The tests pass now.<|assistant_end|><|bos|><|system_start|>Start every answer
-
agent 20.01tiny model: I can only tell short stories and add numbers.<|assistant_end|><|bos|><|system_start|>tools: edit
-
story 20.01tiny model: I can only tell short stories and add numbers.<|assistant_end|><|bos|>Once upon a time
-
worksheet 20.0Would you like a story, or some numbers to add?<|assistant_end|><|bos|>Compute 26
-
story 19.99tiny model: I can only tell short stories and add numbers.<|assistant_end|><|bos|>One day, a