${score_id}

---Question---
${question}

---Answer to Evaluate---
${answer}
---End of Answer---

---Global Assertion to Check---
${assertion}

---Supporting Assertions---
The following assertions are local/supporting assertions derived from source documents:
${supporting_assertions}

---Instructions---
Follow these steps in order:

1. **Global Assertion (FIRST)**: Evaluate whether the answer satisfies the global assertion **on its own merits**, exactly as you would evaluate any standalone assertion. If the answer conveys semantically equivalent information that satisfies the global assertion, it passes. Do this BEFORE considering the supporting assertions.

2. **Supporting Assertions (for coverage tracking)**: For each supporting assertion (SA1, SA2, etc.), determine if the answer contains evidence that satisfies it. Provide reasoning for each, citing specific evidence from the answer for why the assertion passed or failed. These results track what specific evidence is present but do NOT change your global assertion result from Step 1.

3. **Discovery Detection**: Identify if the answer contains factual information that (a) is NOT covered by any supporting assertion, AND (b) helps satisfy the global assertion. This indicates the answer contains additional relevant information beyond the provided supporting assertions.

**Important Notes:**
- For both supporting and global assertions, semantic equivalence counts - if the answer conveys the same meaning using different words, it satisfies the assertion
- The supporting assertions track coverage; they do NOT define what evidence is required for the global assertion
- Discovery must be relevant to satisfying the global assertion, not just any extra details
- Return results for each supporting assertion using its ID (SA1, SA2, etc.)

Format your response as a JSON object (use 1 for true, 0 for false):
{
    "reasoning": "Overall explanation of the global assertion evaluation, citing specific evidence from the answer.",
    "global_passed": 1/0,
    "supporting_results": [
        {"id": "SA1", "reasoning": "Explanation citing specific evidence for why the assertion passed or failed", "passed": 1/0},
        {"id": "SA2", "reasoning": "Explanation citing specific evidence for why the assertion passed or failed", "passed": 1/0},
        ...
    ],
    "discovery_reasoning": "Explanation of discovery or empty string",
    "has_discovery": 1/0
}
