Feature #1024

Strongest examples

On text the dictionary never trained on. The underlined token is where it fires most.

1. agent 2.42
   <|assistant_end|><|tool_start|>Edited number_utils.py (1 replacement)<|tool_end|><|assistant_start|>
2. agent 2.42
   <|assistant_end|><|tool_start|>Edited number_utils.py (1 replacement)<|tool_end|><|assistant_start|>
3. agent 2.39
   for n in nums]<|tool_call_end|><|assistant_end|><|tool_start|>Edited stats.py (1 replacement)<|tool_end|><|assistant_start|>
4. agent 2.35
   s + "!"<|tool_call_end|><|assistant_end|><|tool_start|>Edited greetings.py (1 replacement)<|tool_end|><|assistant_start|>
5. agent 2.33
   <|assistant_end|><|tool_start|>Edited main.py (1 replacement)<|tool_end|><|assistant_start|>
6. agent 2.33
   <|assistant_end|><|tool_start|>Edited main.py (1 replacement)<|tool_end|><|assistant_start|>
7. chat 2.24
   how to tell little stories and how to add numbers. Try asking me for a story!<|assistant_end|>
8. chat 2.23
   how to tell little stories and how to add numbers. Try asking me for a story!<|assistant_end|>
9. agent 2.23
   x + 1<|tool_call_end|><|assistant_end|><|tool_start|>Edited math_utils.py (1 replacement)<|tool_end|><|assistant_start|>
10. chat 2.22
    how to tell little stories and how to add numbers. Try asking me for a story!<|assistant_end|>
11. agent 2.21
    <|assistant_end|><|tool_start|>Edited arith.py (1 replacement)<|tool_end|><|assistant_start|>
12. agent 2.21
    (nums)<|tool_call_end|><|assistant_end|><|tool_start|>Edited listlib.py (1 replacement)<|tool_end|><|assistant_start|>