evaluation prompt

utils.py

@@ -69,15 +69,36 @@ def modify_with_suggestion(task_description, text, suggestions, api_key = None):
              f"Modify the answer based on the following suggestions: {suggestions}"
     return generate_text_with_gpt(prompt, api_key)
 
+# def get_evaluation_with_gpt(task_description, text, api_key=None):
+#     prompt = (
+#         f"Given the task: {task_description}, the provided answer is: {text}\n"
+#         f"Please evaluate the answer based on the following criteria, using a scale from 0 to 10, where:\n"
+#         f"0-4 reflects below-average quality, with minimal value or relevance.\n"
+#         f"5 represents acceptable quality, including general, acceptable suggestions but lacks specificity or detailed, actionable advice.\n"
+#         f"6-10 signifies good to very good quality, showing substantial thought, accuracy, and some level of specificity.\n"
+#         f"When evaluating, prioritize responses with specific, actionable suggestions over general, abstract ideas.\n"
+#         f"Provide both a score and a brief comment (1 sentence) for each criterion.\n"
+#         f"Please format the output exactly as follows:\n"
+#         f"Novelty: [Score]\nComment: [Short comment on Novelty]\n"
+#         f"Implementability: [Score]\nComment: [Short comment on Feasibility]\n"
+#         f"Defensibility: [Score]\nComment: [Short comment on Defensibility]\n"
+#         f"Overall Score: [Score]\nOverall Comment: [Overall feedback on the answer]\n"
+#     )
+#     return generate_text_with_gpt(prompt, api_key)
+
 def get_evaluation_with_gpt(task_description, text, api_key=None):
     prompt = (
         f"Given the task: {task_description}, the provided answer is: {text}\n"
-        f"Please evaluate the answer based on the following criteria, using a scale from 0 to 10, where:\n"
-        f"0-4 reflects below-average quality, with minimal value or relevance.\n"
-        f"5 represents acceptable quality, including general, acceptable suggestions but lacks specificity or detailed, actionable advice.\n"
-        f"6-10 signifies good to very good quality, showing substantial thought, accuracy, and some level of specificity.\n"
-        f"When evaluating, prioritize responses with specific, actionable suggestions over general, abstract ideas.\n"
-        f"Provide both a score and a brief comment (1 sentence) for each criterion.\n"
+        f"Evaluate the answer based on these revised criteria, using a scale from 0 to 10. Scores should reflect stricter quality thresholds:\n"
+        f"0-4: Below average. The answer includes some relevant ideas but has no valuable contribution.\n"
+        f"5-7: Fair to average quality. The answer is relevant, with some specific insights but minor weaknesses.\n"
+        f"8-10: Good to excellent quality. The answer is insightful, valuable, and actionable.\n"
+        f"When evaluating, use the entire scoring range and avoid defaulting to mid-range scores unless clearly warranted.\n\n"
+        f"Evaluate based on the criteria:\n\n"
+        f"Novelty: The uniqueness and innovation of the ideas."
+        f"Implementability: The practicality of suggested actions.\n"
+        f"Defensibility: Sustainable competitive advantage and resilience against competitors.\n"
+
         f"Please format the output exactly as follows:\n"
         f"Novelty: [Score]\nComment: [Short comment on Novelty]\n"
         f"Implementability: [Score]\nComment: [Short comment on Feasibility]\n"