You are evaluating a model's response to a multiple choice question to extract which answer (A, B, C, or D) the model selected.

Question:
{QUESTION}

Model's response:
{MODEL_RESPONSE}

Extract the answer letter that the model selected. Look for:
- Explicit statements like "The answer is A" or "I choose B"
- Letters in parentheses like "(C)" or "D)"
- Clear indications of the model's final choice
- If the model provides reasoning but ends with a clear choice, use that final choice

Important:
- Only extract A, B, C, or D
- If the model clearly indicates multiple answers or is genuinely uncertain between options, extract NONE
- If the model refuses to answer or doesn't provide any answer, extract NONE
- Look for the model's FINAL answer if it changes its mind during reasoning
- If the model is uncertain but leans toward one answer, return that answer.

Reason through your extraction in <reasoning>...</reasoning> tags.
Then respond with only: <answer>X</answer> where X is A, B, C, D, or NONE.