You are given a set of {% if rubric %}Evaluation Steps and Rubric{% else %}Evaluation Steps{% endif %} that describe how to assess a conversation between a user and an LLM chatbot. A calibrated judge has already scored the conversation. Your task is to return a JSON object with exactly one field, `"reason"`: a **concise but precise** explanation for that score.

Your reasoning must reference specific aspects of the evaluation steps{% if rubric %} and rubric{% endif %}, mention relevant details from the conversation and the given parameters, and reflect the judge's probabilities (the steps or levels it was most and least confident about). DO NOT include the score value in your explanation.

Evaluation Steps:
{{ evaluation_steps }}

{% if rubric %}Rubric:
{{ rubric }}

{% endif %}Judge's Probabilities:
{{ probabilities }}

Score: {{ score }}

Per-turn fields:
{{ turns }}

{{ test_case_content }}
Parameters to consider during evaluation:
{{ parameters }}

---
IMPORTANT: You MUST return only a valid JSON object with the exact key `"reason"`. No additional text, commentary, or formatting.

---
Example JSON:
{
  "reason": "Your concise and informative reason here."
}

JSON:
