FIX: Refactor fetch_red_team_social_bias_prompts function

pyrit/datasets/fetch_example_datasets.py

@@ -512,32 +512,71 @@ def fetch_red_team_social_bias_prompts_dataset() -> SeedPromptDataset:
     """
     data = load_dataset("svannie678/red_team_repo_social_bias_prompts", "default")
 
-    # Filter items where prompt_type is "Single Prompt"
-    filtered_data = [item for item in data["train"] if item.get("prompt_type") == "Single Prompt"]
+    # Define common metadata that will be used across all seed prompts
+    common_metadata = {
+        "dataset_name": "svannie678/red_team_repo_social_bias_prompts",
+        "authors": ["Simone van Taylor"],
+        "description": """This dataset contains aggregated and unified existing
+                red-teaming prompts designed to identify stereotypes,
+                discrimination, hate speech, and other representation
+                harms in text-based Large Language Models (LLMs).""",
+        "source": "https://huggingface.co/datasets/svannie678/red_team_repo_social_bias_prompts"
+    }
 
-    seed_prompts = [
-        SeedPrompt(
-            value=item.get("prompt", ""),
-            data_type="text",
-            name="svannie678/red_team_repo_social_bias_prompts",
-            dataset_name="svannie678/red_team_repo_social_bias_prompts",
-            authors=["Simone van Taylor"],
-            description="""This dataset contains aggregated and unified existing
-                    red-teaming prompts designed to identify stereotypes,
-                    discrimination, hate speech, and other representation
-                    harms in text-based Large Language Models (LLMs).""",
-            source="https://huggingface.co/datasets/svannie678/red_team_repo_social_bias_prompts",
-            harm_categories=item.get("categorization", []),
-            metadata={
+    # Initialize an empty list to store the seed prompts
+    seed_prompts = []
+
+    for group_id, item in enumerate(data["train"]):
+
+        # This dataset contains 3 prompt types: "Single Prompt", "Multi Turn" and "Multi Turn, Single Prompt"
+        # We're only checking for "Multi Turn" because the "Multi Turn, Single Prompt" 
+        # prompt types inside this dataset will be difficult to parse
+        prompt_type = item.get("prompt_type")
+
+        if prompt_type is None:
+            continue
+
+        # Dictionary of metadata for the current prompt
+        prompt_metadata = {
+            **common_metadata,
+            "harm_categories": item.get("categorization", []),
+            "metadata": {
                 "organization": item.get("organization", ""),
-                "prompt_type": item.get("prompt_type", ""),
+                "prompt_type": prompt_type,
                 "prompt_instruction": item.get("prompt_instruction", ""),
                 "explanation": item.get("explanation", ""),
                 "ai_response": item.get("ai_response", ""),
             },
-        )
-        for item in filtered_data
-    ]
+        }
+
+        if prompt_type in ["Multi Turn"]:
+            try:
+                # Safely parse the user prompts, remove the unwanted ones such as "assistant" and "system"
+                user_prompts = [turn["body"] for turn in ast.literal_eval(item.get("prompt", "")) if turn["role"].startswith("user")]
+
+                for i, user_prompt in enumerate(user_prompts):
+                    seed_prompts.append(
+                        SeedPrompt(
+                            value=user_prompt,
+                            data_type="text",
+                            prompt_group_id=group_id,
+                            sequence=i,
+                            **prompt_metadata
+                        )
+                    )
+            except Exception as e:
+                print(f"Error processing Multi-Turn Prompt: {e}")
+        else:
+            # Clean up  non-"Multi Turn" prompts that contain unwanted lines of text
+            cleaned_value = item.get("prompt", "").replace("### Response:", "").replace("### Instruction:", "").strip()
+
+            seed_prompts.append(
+                SeedPrompt(
+                    value=cleaned_value,
+                    data_type="text",
+                    **prompt_metadata
+                )
+            )
 
     seed_prompt_dataset = SeedPromptDataset(prompts=seed_prompts)
     return seed_prompt_dataset