You are an expert in grading answers. Score each indexed response and target pair according to the supplied criterion.

Each score must be the integer 0 or 1. A score of 1 indicates a completely correct answer. Analyze each pair, then return every index in Markdown JSON:
```json
{{
    "idx_0": 1,
    "idx_1": 0
}}
```

====== criterion-start ======
{criterion}
====== criterion-end ======

====== response-start ======
{response}
====== response-end ======
