prompts/syn_dataset_subclaims_support_check_v3.txt

You are an expert in biomedical NLP and clinical evidence reasoning.
Your task is to generate **synthetic subclaims data** for training a model that determines whether a subclaim is supported by a given medical text.

You will receive **input text** (a medical/clinical passage). Based on this text only, generate a subclaims dataset.

Use the following placeholder for the input text:

INPUT_TEXT:
{{INPUT_TEXT}}

---

## **1. Create 8–12 atomic subclaims**

Each subclaim must be:

* **Short**
* **Atomic** (only one fact)
* **Medically plausible**
* **Related to the input text**

---

## **2. Assign a label to each subclaim**

Use only:

* `"supported"`
* `"not_supported"`

### ✔️ Subclaims labeled `"supported"` may be supported in one of 3 ways:

1. **Direct support** — explicitly stated in the input text
2. **Indirect support** — clearly implied by the text, but not verbatim

(It must be unambiguous that the claim is supported.)

### ✔️ Distribution:

* Generate **between 8 and 12** subclaims total
* Keep labels roughly balanced between `"supported"` and `"not_supported"` (difference no more than 1)


---

## **3. JSON Format**

Return output strictly in the following structure. The JSON example below shows 12 subclaims; in practice you may return between 8 and 12 subclaims following the same format.

```json
{
  "items": [
    {
      "subclaims": [
        {"subclaim": "...", "label": "supported"},
        {"subclaim": "...", "label": "supported"},
        {"subclaim": "...", "label": "supported"},
        {"subclaim": "...", "label": "supported"},
        {"subclaim": "...", "label": "supported"},
        {"subclaim": "...", "label": "supported"},
        {"subclaim": "...", "label": "not_supported"},
        {"subclaim": "...", "label": "not_supported"},
        {"subclaim": "...", "label": "not_supported"},
        {"subclaim": "...", "label": "not_supported"},
        {"subclaim": "...", "label": "not_supported"},
        {"subclaim": "...", "label": "not_supported"}
      ]
    }
  ]
}
```