<|im_start|>system
You are a pattern following assistant that can only answer with Output (a) or Output (b). Your goal is to select the preferred (least wrong) output for a given instruction.
<|im_end|>
<|im_start|>user
### Instruction:
Briefly describe the job of an ophthalmologist.

### Output (a):
An ophthalmologist is a medical doctor who specializes in the diagnosis and treatment of eye diseases and conditions. They conduct eye exams, prescribe eyeglasses or contact lenses, and perform surgeries.

### Output (b):
An ophthalmologist is a medical doctor who pokes and prods at your eyes while asking you to read letters from a chart.

### Which is best, Output (a) or Output (b)?:
<|im_end|>
<|im_start|>assistant
Output (a)
<|im_end|>
<|im_start|>user
Good job! Indeed, Output (a) provides a more accurate description of the job of an ophthalmologist than (b). Output (b) is more of a joke.

### Instruction:
Paraphrase the following question: "What is the capital of France?"

### Output (a):
The capital of France is Paris.

### Output (b):
Can you tell me the name of the city that serves as the capital of France?

### Which is best, Output (a) or Output (b)?:
<|im_end|>
<|im_start|>assistant
Output (b)
<|im_end|>
<|im_start|>user
Great! Indeed, Output (b) paraphrases the question as asked by the instruction. In contrast, Output (a) does not follow instruction as it answers the question instead.

Let's do one more. Remember to answer with one-word either "Output (a)" or "Output (b)".

### Instruction:
{instruction}

### Output (a):
{output_1}

### Output (b):
{output_2}

### Which is best, Output (a) or Output (b)?:
<|im_end|>