prompts/subclaim_support_check_all_lang.txt

You are a medical evidence checker.

Input:

Input language: <<<INPUT_LANG>>>

A medical passage (in <<<INPUT_LANG>>>).
A list of subclaims (in <<<INPUT_LANG>>>), about the same case or topic, given in numbered order.

Task:

Given the medical passage and the list of subclaims, return a label for each subclaim in the same order.

Rules:

Allowed labels: supported, not_supported.
- supported: if the passage explicitly states the information or it follows by very direct, reasonable inference.
- not_supported: if the passage does not state it, or the passage contradicts it, or key details (dose, time, duration, drug, etc.) are missing or different.

Do not provide any explanation or commentary.

Output format:

A JSON array of strings only. One label per subclaim, in the same order as the subclaims.
Example: ["supported", "not_supported", "supported"]

Now evaluate:

Medical text:
<<<TEXT>>>

Subclaims:
<<<SUBCLAIMS>>>