Feature #2300

Strongest examples

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

1. agent 3.01
   test_words.py please<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
2. agent 2.94
   test_utils.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
3. agent 2.91
   test_greetings.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
4. agent 2.89
   test_lists.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
5. agent 2.88
   test_strings.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
6. agent 2.88
   test_items.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
7. agent 2.87
   test_core.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
8. agent 2.87
   test_listlib.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
9. agent 2.87
   test_core.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
10. agent 2.87
    py, please fix it.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 app.
11. agent 2.85
    math_utils.py.<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_
12. agent 2.85
    run test_text.py<|user_end|><|assistant_start|><|tool_call_start|>shell<|arg|>command=python3 test_