Fix REIMB_TERM being converted to list in lesser_of_distribution

- Fixed prompt_lesser_of_distribution to extract string from list when parser returns list
- Added debug print statements to trace REIMB_TERM format through processing pipeline
- Removed defensive normalization checks that are no longer needed upstream
- Added debug output in reimbursement_level and methodology_breakout_single_row to track data flow

src/pipelines/clients/bcbs_promise/prompts/prompt_calls.py

@@ -119,6 +119,32 @@ def prompt_reimbursement_primary(
 
     try:
         llm_answer_final = _parser(llm_answer_raw)
+        # Normalize SERVICE_TERM and REIMB_TERM to strings immediately after parsing
+        # These should always be strings per row, not lists
+        if isinstance(llm_answer_final, list):
+            for answer_dict in llm_answer_final:
+                if isinstance(answer_dict, dict):
+                    # SERVICE_TERM should be a string per row
+                    if "SERVICE_TERM" in answer_dict:
+                        service_term = answer_dict["SERVICE_TERM"]
+                        if isinstance(service_term, list):
+                            # If list, join with comma (shouldn't happen, but defensive)
+                            answer_dict["SERVICE_TERM"] = ", ".join(str(item) for item in service_term if item)
+                        elif service_term is not None:
+                            answer_dict["SERVICE_TERM"] = str(service_term)
+                        else:
+                            answer_dict["SERVICE_TERM"] = ""
+                    
+                    # REIMB_TERM should be a string per row
+                    if "REIMB_TERM" in answer_dict:
+                        reimb_term = answer_dict["REIMB_TERM"]
+                        if isinstance(reimb_term, list):
+                            # If list, join with comma (shouldn't happen, but defensive)
+                            answer_dict["REIMB_TERM"] = ", ".join(str(item) for item in reimb_term if item)
+                        elif reimb_term is not None:
+                            answer_dict["REIMB_TERM"] = str(reimb_term)
+                        else:
+                            answer_dict["REIMB_TERM"] = ""
     except ValueError as e:
         logging.error(f"Error parsing LLM response: {e}")
         logging.error(f"Raw LLM output: {llm_answer_raw}")
@@ -151,6 +177,19 @@ def prompt_methodology_breakout(
     logging.debug(f"LLM Response for {filename}: {llm_response}")
     try:
         methodology_breakout_answers = _parser(llm_response)
+        # Normalize single-value fields that should be strings (not lists)
+        # FEE_SCHEDULE should be a string per methodology breakout result
+        if isinstance(methodology_breakout_answers, list):
+            for methodology_dict in methodology_breakout_answers:
+                if isinstance(methodology_dict, dict) and "FEE_SCHEDULE" in methodology_dict:
+                    fee_schedule = methodology_dict["FEE_SCHEDULE"]
+                    if isinstance(fee_schedule, list):
+                        # If list, take first element (shouldn't happen, but defensive)
+                        methodology_dict["FEE_SCHEDULE"] = fee_schedule[0] if fee_schedule else ""
+                    elif fee_schedule is not None:
+                        methodology_dict["FEE_SCHEDULE"] = str(fee_schedule)
+                    else:
+                        methodology_dict["FEE_SCHEDULE"] = ""
     except ValueError as e:
         logging.error(f"Error parsing LLM response: {e}")
         methodology_breakout_answers = []
@@ -166,9 +205,16 @@ def prompt_fee_schedule_breakout(
     Call FEE_SCHEDULE_BREAKOUT prompt with cached instruction.
     Note: Field definitions are now included in FEE_SCHEDULE_BREAKOUT_INSTRUCTION() for caching.
     """
+    # Normalize FEE_SCHEDULE to string before using in prompt (should already be normalized from prompt_methodology_breakout)
+    fee_schedule = methodology_breakout_dict.get("FEE_SCHEDULE", "")
+    if isinstance(fee_schedule, list):
+        # If list, take first element (shouldn't happen, but defensive)
+        fee_schedule = fee_schedule[0] if fee_schedule else ""
+    fee_schedule = str(fee_schedule) if fee_schedule else ""
+    
     prompt, _parser = prompt_templates.FEE_SCHEDULE_BREAKOUT(
         reimbursement_method,
-        methodology_breakout_dict.get("FEE_SCHEDULE"),
+        fee_schedule,
     )
     logging.debug(f"Fee Schedule Breakout Prompt for {filename}: {prompt}")
     llm_answer_raw = llm_utils.invoke_claude(
@@ -299,7 +345,13 @@ def prompt_lob_relationship(
         instruction=prompt_templates.LOB_RELATIONSHIP_INSTRUCTION(),
     )
     llm_answer_final = _parser(llm_answer_raw)
-    return llm_answer_final
+    # LOB_RELATIONSHIP returns a list (e.g., ["Inclusive"]), but we need a string
+    if isinstance(llm_answer_final, list) and len(llm_answer_final) > 0:
+        return str(llm_answer_final[0])
+    elif isinstance(llm_answer_final, str):
+        return llm_answer_final
+    else:
+        return "Exclusive"  # Default fallback
 
 
 def prompt_special_case_assignment(

src/pipelines/clients/clover/prompts/prompt_calls.py

@@ -126,6 +126,32 @@ def prompt_reimbursement_primary(
 
     try:
         llm_answer_final = _parser(llm_answer_raw)
+        # Normalize SERVICE_TERM and REIMB_TERM to strings immediately after parsing
+        # These should always be strings per row, not lists
+        if isinstance(llm_answer_final, list):
+            for answer_dict in llm_answer_final:
+                if isinstance(answer_dict, dict):
+                    # SERVICE_TERM should be a string per row
+                    if "SERVICE_TERM" in answer_dict:
+                        service_term = answer_dict["SERVICE_TERM"]
+                        if isinstance(service_term, list):
+                            # If list, join with comma (shouldn't happen, but defensive)
+                            answer_dict["SERVICE_TERM"] = ", ".join(str(item) for item in service_term if item)
+                        elif service_term is not None:
+                            answer_dict["SERVICE_TERM"] = str(service_term)
+                        else:
+                            answer_dict["SERVICE_TERM"] = ""
+                    
+                    # REIMB_TERM should be a string per row
+                    if "REIMB_TERM" in answer_dict:
+                        reimb_term = answer_dict["REIMB_TERM"]
+                        if isinstance(reimb_term, list):
+                            # If list, join with comma (shouldn't happen, but defensive)
+                            answer_dict["REIMB_TERM"] = ", ".join(str(item) for item in reimb_term if item)
+                        elif reimb_term is not None:
+                            answer_dict["REIMB_TERM"] = str(reimb_term)
+                        else:
+                            answer_dict["REIMB_TERM"] = ""
     except ValueError as e:
         logging.error(f"Error parsing LLM response: {e}")
         logging.error(f"Raw LLM output: {llm_answer_raw}")
@@ -158,6 +184,19 @@ def prompt_methodology_breakout(
     logging.debug(f"LLM Response for {filename}: {llm_response}")
     try:
         methodology_breakout_answers = _parser(llm_response)
+        # Normalize single-value fields that should be strings (not lists)
+        # FEE_SCHEDULE should be a string per methodology breakout result
+        if isinstance(methodology_breakout_answers, list):
+            for methodology_dict in methodology_breakout_answers:
+                if isinstance(methodology_dict, dict) and "FEE_SCHEDULE" in methodology_dict:
+                    fee_schedule = methodology_dict["FEE_SCHEDULE"]
+                    if isinstance(fee_schedule, list):
+                        # If list, take first element (shouldn't happen, but defensive)
+                        methodology_dict["FEE_SCHEDULE"] = fee_schedule[0] if fee_schedule else ""
+                    elif fee_schedule is not None:
+                        methodology_dict["FEE_SCHEDULE"] = str(fee_schedule)
+                    else:
+                        methodology_dict["FEE_SCHEDULE"] = ""
     except ValueError as e:
         logging.error(f"Error parsing LLM response: {e}")
         methodology_breakout_answers = []
@@ -173,9 +212,16 @@ def prompt_fee_schedule_breakout(
     Call FEE_SCHEDULE_BREAKOUT prompt with cached instruction.
     Note: Field definitions are now included in FEE_SCHEDULE_BREAKOUT_INSTRUCTION() for caching.
     """
+    # Normalize FEE_SCHEDULE to string before using in prompt (should already be normalized from prompt_methodology_breakout)
+    fee_schedule = methodology_breakout_dict.get("FEE_SCHEDULE", "")
+    if isinstance(fee_schedule, list):
+        # If list, take first element (shouldn't happen, but defensive)
+        fee_schedule = fee_schedule[0] if fee_schedule else ""
+    fee_schedule = str(fee_schedule) if fee_schedule else ""
+    
     prompt, _parser = prompt_templates.FEE_SCHEDULE_BREAKOUT(
         reimbursement_method,
-        methodology_breakout_dict.get("FEE_SCHEDULE"),
+        fee_schedule,
     )
     logging.debug(f"Fee Schedule Breakout Prompt for {filename}: {prompt}")
     llm_answer_raw = llm_utils.invoke_claude(
@@ -306,7 +352,13 @@ def prompt_lob_relationship(
         instruction=prompt_templates.LOB_RELATIONSHIP_INSTRUCTION(),
     )
     llm_answer_final = _parser(llm_answer_raw)
-    return llm_answer_final
+    # LOB_RELATIONSHIP returns a list (e.g., ["Inclusive"]), but we need a string
+    if isinstance(llm_answer_final, list) and len(llm_answer_final) > 0:
+        return str(llm_answer_final[0])
+    elif isinstance(llm_answer_final, str):
+        return llm_answer_final
+    else:
+        return "Exclusive"  # Default fallback
 
 
 def prompt_special_case_assignment(

src/pipelines/saas/prompts/prompt_calls.py

@@ -126,6 +126,11 @@ def prompt_reimbursement_primary(
 
     try:
         llm_answer_final = _parser(llm_answer_raw)
+        
+        # DEBUG: Print raw parsed output
+        print(f"\n[DEBUG REIMBURSEMENT_PRIMARY] Parsed output for {filename}:")
+        print(llm_answer_final)
+        print()
     except ValueError as e:
         logging.error(f"Error parsing LLM response: {e}")
         logging.error(f"Raw LLM output: {llm_answer_raw}")
@@ -173,9 +178,11 @@ def prompt_fee_schedule_breakout(
     Call FEE_SCHEDULE_BREAKOUT prompt with cached instruction.
     Note: Field definitions are now included in FEE_SCHEDULE_BREAKOUT_INSTRUCTION() for caching.
     """
+    # Normalize FEE_SCHEDULE to string before using in prompt (should already be normalized from prompt_methodology_breakout)
+    fee_schedule = methodology_breakout_dict.get("FEE_SCHEDULE", "")    
     prompt, _parser = prompt_templates.FEE_SCHEDULE_BREAKOUT(
         reimbursement_method,
-        methodology_breakout_dict.get("FEE_SCHEDULE"),
+        fee_schedule,
     )
     logging.debug(f"Fee Schedule Breakout Prompt for {filename}: {prompt}")
     llm_answer_raw = llm_utils.invoke_claude(
@@ -306,7 +313,13 @@ def prompt_lob_relationship(
         instruction=prompt_templates.LOB_RELATIONSHIP_INSTRUCTION(),
     )
     llm_answer_final = _parser(llm_answer_raw)
-    return llm_answer_final
+    # LOB_RELATIONSHIP returns a list (e.g., ["Inclusive"]), but we need a string
+    if isinstance(llm_answer_final, list) and len(llm_answer_final) > 0:
+        return str(llm_answer_final[0])
+    elif isinstance(llm_answer_final, str):
+        return llm_answer_final
+    else:
+        return "Exclusive"  # Default fallback
 
 
 def prompt_special_case_assignment(
@@ -687,13 +700,17 @@ def prompt_lesser_of_distribution(
         return reimb_term  # Return original term unchanged
     else:
         llm_answer_final = _parser(llm_answer_raw)
+        
+        # DEBUG: Print raw parsed output
+        print(f"\n[DEBUG prompt_lesser_of_distribution] Raw parsed output:")
+        print(llm_answer_final)
 
         logging.debug(
             f"Applied lesser-of to '{service_term}' on page {page_num}: "
             f"{reimb_term} → {llm_answer_final[:60]}..."
         )
 
-        return llm_answer_final
+        return llm_answer_final[0] if isinstance(llm_answer_final, list) else llm_answer_final
 
 
 def prompt_lesser_of_check(

src/pipelines/shared/extraction/dynamic_funcs.py

@@ -234,8 +234,20 @@ def add_one_to_one_field(
     # Only search for these fields when NO LOB has been found
 
     if field_to_add.field_name in ["PROGRAM", "PRODUCT", "NETWORK"]:
-        lob_values = [answer_dict.get("LOB", "") for answer_dict in answer_dicts]
-        unique_lobs = set([lob for lob in lob_values if not string_utils.is_empty(lob)])
+        # LOB can be a string or a list (from JSON format)
+        # Extract all LOB values, handling both string and list formats
+        lob_values = []
+        for answer_dict in answer_dicts:
+            lob_value = answer_dict.get("LOB", "")
+            if isinstance(lob_value, list):
+                # If it's a list, extract each item and ensure it's a string
+                for item in lob_value:
+                    if not string_utils.is_empty(item):
+                        lob_values.append(str(item))
+            elif not string_utils.is_empty(lob_value):
+                # If it's a string, add it directly (already hashable)
+                lob_values.append(lob_value)
+        unique_lobs = set(lob_values)
         if len(unique_lobs) > 0:
             return one_to_one_fields
 
@@ -283,8 +295,22 @@ def get_dynamic_one_to_one_fields(
     # Only search for PROGRAM, PRODUCT, NETWORK when NO LOB has been found
     # NOTE: Check raw LOB field, not AARETE_DERIVED_LOB, because AARETE_DERIVED_LOB
     # is populated later via crosswalk and may be derived from PROGRAM/PRODUCT
-    lob_values = [answer_dict.get("LOB", "") for answer_dict in answer_dicts]
-    unique_lobs = set([lob for lob in lob_values if not string_utils.is_empty(lob)])
+    # LOB can be a string or a list (from JSON format)
+    # Extract all LOB values, handling both string and list formats
+    # All values must be hashable (strings) to create a set
+    lob_values = []
+    for answer_dict in answer_dicts:
+        lob_value = answer_dict.get("LOB", "")
+        if isinstance(lob_value, list):
+            # If it's a list, extract each item and ensure it's a string
+            for item in lob_value:
+                if not string_utils.is_empty(item):
+                    # Convert to string to ensure hashability (handles edge cases)
+                    lob_values.append(str(item))
+        elif not string_utils.is_empty(lob_value):
+            # If it's a string, add it directly (already hashable)
+            lob_values.append(lob_value)
+    unique_lobs = set(lob_values)
     has_lob = len(unique_lobs) > 0
 
     # Handle ALL empty fields

src/pipelines/shared/extraction/one_to_n_funcs.py

@@ -98,6 +98,13 @@ def reimbursement_level(
             reimbursement_primary_answers, constants, filename
         )
 
+    # ============================================================================
+    # DEBUG: Print reimbursement_primary_answers after filtering
+    # ============================================================================
+    print(f"\n[DEBUG reimbursement_level] After filtering (line 101):")
+    print(reimbursement_primary_answers)
+    # ============================================================================
+
     return reimbursement_primary_answers
 
 
@@ -108,10 +115,11 @@ def clean_reimbursement_primary(
 ) -> list[dict[str, str]]:
     """
     Cleans and processes reimbursement primary answers through several steps:
-    1. Split compound reimbursements
-    2. Filter out any non-reimbursement content from split answers
-    3. Deduplicate against previously seen (SERVICE_TERM, REIMB_TERM) pairs
-    4. Apply exhibit lesser-of statement if it exists
+    1. Normalize SERVICE_TERM and REIMB_TERM to strings (they should never be lists per-row)
+    2. Split compound reimbursements
+    3. Filter out any non-reimbursement content from split answers
+    4. Deduplicate against previously seen (SERVICE_TERM, REIMB_TERM) pairs
+    5. Apply exhibit lesser-of statement if it exists
 
     Args:
         reimbursement_primary_answers (list[str]): List of dictionaries containing reimbursement details.
@@ -122,6 +130,7 @@ def clean_reimbursement_primary(
     """
     if reimbursement_primary_answers:  # If any reimbursements found
 
+
         # If needed, add Reimbursement Splitting here
 
         # If needed, add deduplication here
@@ -330,9 +339,17 @@ def methodology_breakout_single_row(
         list[dict]: Updated list with methodology breakout details and fee schedule information.
     """
 
-    reimb_term, service_term = answer_dict.get("REIMB_TERM"), answer_dict.get(
-        "SERVICE_TERM"
-    )
+    # DEBUG: Check what answer_dict contains
+    print(f"\n[DEBUG methodology_breakout_single_row] answer_dict REIMB_TERM: {repr(answer_dict.get('REIMB_TERM'))} (type: {type(answer_dict.get('REIMB_TERM')).__name__})")
+    print(f"[DEBUG methodology_breakout_single_row] answer_dict SERVICE_TERM: {repr(answer_dict.get('SERVICE_TERM'))} (type: {type(answer_dict.get('SERVICE_TERM')).__name__})")
+    
+    reimb_term = answer_dict.get("REIMB_TERM", "")
+    service_term = answer_dict.get("SERVICE_TERM", "")
+    
+    # DEBUG: Check what we extracted
+    print(f"[DEBUG methodology_breakout_single_row] Extracted reimb_term: {repr(reimb_term)} (type: {type(reimb_term).__name__})")
+    print(f"[DEBUG methodology_breakout_single_row] Extracted service_term: {repr(service_term)} (type: {type(service_term).__name__})\n")
+    
     methodology_breakout_questions = FieldSet(
         config.FIELD_JSON_PATH, field_type="methodology_breakout"
     )
@@ -781,6 +798,31 @@ def one_to_n_cleaning(
     ################################ Split REIMB_DATES ################################
     all_exhibit_rows = split_reimb_dates(all_exhibit_rows, filename)
 
+    ################################ Normalize Single-Value Fields to Strings ################################
+    # REIMB_TERM, SERVICE_TERM, and LOB_PROGRAM_RELATIONSHIP should always be strings (not lists)
+    for answer_dict in all_exhibit_rows:
+        # LOB_PROGRAM_RELATIONSHIP should be a string (not a list)
+        if "LOB_PROGRAM_RELATIONSHIP" in answer_dict:
+            lob_program_rel = answer_dict["LOB_PROGRAM_RELATIONSHIP"]
+            if isinstance(lob_program_rel, list):
+                # If list, take first element (shouldn't happen, but defensive)
+                answer_dict["LOB_PROGRAM_RELATIONSHIP"] = str(lob_program_rel[0]) if lob_program_rel else "Exclusive"
+            elif lob_program_rel is not None:
+                answer_dict["LOB_PROGRAM_RELATIONSHIP"] = str(lob_program_rel)
+            else:
+                answer_dict["LOB_PROGRAM_RELATIONSHIP"] = "Exclusive"
+        
+        # LOB_PRODUCT_RELATIONSHIP should be a string (not a list)
+        if "LOB_PRODUCT_RELATIONSHIP" in answer_dict:
+            lob_product_rel = answer_dict["LOB_PRODUCT_RELATIONSHIP"]
+            if isinstance(lob_product_rel, list):
+                # If list, take first element (shouldn't happen, but defensive)
+                answer_dict["LOB_PRODUCT_RELATIONSHIP"] = str(lob_product_rel[0]) if lob_product_rel else "Exclusive"
+            elif lob_product_rel is not None:
+                answer_dict["LOB_PRODUCT_RELATIONSHIP"] = str(lob_product_rel)
+            else:
+                answer_dict["LOB_PRODUCT_RELATIONSHIP"] = "Exclusive"
+
     return all_exhibit_rows
 
 

src/pipelines/shared/postprocessing/aarete_derived.py

@@ -32,18 +32,26 @@ def fill_na_mapping(answer_dicts):
         all_lob_values = set()
 
         # Get existing AARETE_DERIVED_LOB values (if any)
+        # Handle both list and string formats (JSON lists or pipe-delimited strings)
         existing_lob_list = answer_dict.get("AARETE_DERIVED_LOB", "")
 
         if not string_utils.is_empty(existing_lob_list):
-            for lob_val in existing_lob_list:
+            # Normalize to list format
+            if isinstance(existing_lob_list, list):
+                lob_items = existing_lob_list
+            elif isinstance(existing_lob_list, str):
+                # Handle pipe-delimited format (backward compatibility)
                 if "|" in existing_lob_list:
-                    all_lob_values.update(
-                        v.strip()
-                        for v in lob_val.split("|")
-                        if v.strip() and v.strip() != "N/A"
-                    )
-                elif lob_val.strip() and lob_val.strip() != "N/A":
-                    all_lob_values.add(lob_val)
+                    lob_items = [v.strip() for v in existing_lob_list.split("|")]
+                else:
+                    lob_items = [existing_lob_list]
+            else:
+                lob_items = [str(existing_lob_list)]
+
+            # Process each LOB value
+            for lob_val in lob_items:
+                if isinstance(lob_val, str) and lob_val.strip() and lob_val.strip() != "N/A":
+                    all_lob_values.add(lob_val.strip())
 
         # Get AARETE_DERIVED_LOB from AARETE_DERIVED_PROGRAM crosswalk (always check if PROGRAM exists)
         program_lob_values = set()
@@ -58,16 +66,24 @@ def fill_na_mapping(answer_dicts):
                 not string_utils.is_empty(program_filled_value_list)
                 and "N/A" not in program_filled_value_list
             ):
-                for program_lob_val in program_filled_value_list:
-                    if "|" in program_lob_val:
-                        program_lob_values = set(
-                            v.strip()
-                            for v in program_lob_val.split("|")
-                            if v.strip() and v.strip() != "N/A"
-                        )
-                    elif program_lob_val.strip() and program_lob_val.strip() != "N/A":
-                        program_lob_values = {program_lob_val.strip()}
-                    all_lob_values.update(program_lob_values)
+                # Normalize to list format (handle both list and string)
+                if isinstance(program_filled_value_list, list):
+                    program_lob_items = program_filled_value_list
+                elif isinstance(program_filled_value_list, str):
+                    # Handle pipe-delimited format (backward compatibility)
+                    if "|" in program_filled_value_list:
+                        program_lob_items = [v.strip() for v in program_filled_value_list.split("|")]
+                    else:
+                        program_lob_items = [program_filled_value_list]
+                else:
+                    program_lob_items = [str(program_filled_value_list)]
+
+                # Process each program LOB value
+                for program_lob_val in program_lob_items:
+                    if isinstance(program_lob_val, str) and program_lob_val.strip() and program_lob_val.strip() != "N/A":
+                        program_lob_values.add(program_lob_val.strip())
+                # Update all_lob_values after processing all program LOB values
+                all_lob_values.update(program_lob_values)
 
         # Get AARETE_DERIVED_LOB from PRODUCT crosswalk (always check if PRODUCT exists)
         product_lob_values = set()
@@ -82,16 +98,24 @@ def fill_na_mapping(answer_dicts):
                 not string_utils.is_empty(product_filled_value_list)
                 and "N/A" not in product_filled_value_list
             ):
-                for product_lob_val in product_filled_value_list:
-                    if "|" in product_lob_val:
-                        product_lob_values = set(
-                            v.strip()
-                            for v in product_filled_value_list.split("|")
-                            if v.strip() and v.strip() != "N/A"
-                        )
-                    elif product_lob_val.strip() and product_lob_val.strip() != "N/A":
-                        product_lob_values = {product_lob_val.strip()}
-                    all_lob_values.update(product_lob_values)
+                # Normalize to list format (handle both list and string)
+                if isinstance(product_filled_value_list, list):
+                    product_lob_items = product_filled_value_list
+                elif isinstance(product_filled_value_list, str):
+                    # Handle pipe-delimited format (backward compatibility)
+                    if "|" in product_filled_value_list:
+                        product_lob_items = [v.strip() for v in product_filled_value_list.split("|")]
+                    else:
+                        product_lob_items = [product_filled_value_list]
+                else:
+                    product_lob_items = [str(product_filled_value_list)]
+
+                # Process each product LOB value
+                for product_lob_val in product_lob_items:
+                    if isinstance(product_lob_val, str) and product_lob_val.strip() and product_lob_val.strip() != "N/A":
+                        product_lob_values.add(product_lob_val.strip())
+                # Update all_lob_values after processing all product LOB values
+                all_lob_values.update(product_lob_values)
 
         # Set the merged, deduplicated value
         if all_lob_values:
@@ -165,11 +189,16 @@ def get_crosswalk_fields(answer_dicts: list, constants: Constants):
                             if "AARETE_DERIVED" in to_field_name:
                                 to_field_answer_list.append(individual_from_field_value)
                             else:
-                                to_field_answer_list.append(
-                                    crosswalk.create_reverse_mapping().get(
-                                        individual_from_field_value
-                                    )
-                                )
+                                # Reverse mapping returns dict[str, list[str]], so .get() returns a list
+                                # We need to extend, not append, to avoid nested lists
+                                reverse_mapping = crosswalk.create_reverse_mapping()
+                                reverse_result = reverse_mapping.get(individual_from_field_value)
+                                if reverse_result is not None:
+                                    # reverse_result is a list, extend it
+                                    to_field_answer_list.extend(reverse_result)
+                                else:
+                                    # If not found in reverse mapping, keep original value
+                                    to_field_answer_list.append(individual_from_field_value)
                     answer_dict[to_field_name] = to_field_answer_list
                     # update from field value to list if not already
                     if not isinstance(from_field_value, list):

src/prompts/prompt_templates.py

@@ -26,7 +26,8 @@ JSON_LIST_OF_DICTS_FORMAT_INSTRUCTIONS = """Return your final answer as a valid
 - Each item in the list should be a dictionary with the specified field names as keys.
 - If multiple items are found, include all of them as separate dictionaries in the list.
 - Use "N/A" for fields within each dictionary where the information doesn't apply.
-- Ensure the JSON is properly formatted with square brackets, curly braces, and double quotes."""
+- Ensure the JSON is properly formatted with square brackets, curly braces, and double quotes.
+- The Keys and Values must be strings, and not lists."""
 
 
 # Parser helper functions for prompt templates
@@ -486,7 +487,7 @@ def REIMBURSEMENT_PRIMARY(context) -> Tuple[str, Callable[[str], list]]:
     Call REIMBURSEMENT_PRIMARY_INSTRUCTION() separately for the cached instruction.
 
     Returns:
-        Tuple of (prompt_string, parser_function) where parser expects JSON list output.
+        Tuple of (prompt_string, parser_function) where parser expects JSON list output. Should return a list of dicts of strings
     """
     prompt = f"""
 [START CONTEXT]
@@ -500,7 +501,7 @@ Briefly talk through your reasoning. Then return a properly formatted JSON list
 
 
 def REIMBURSEMENT_PRIMARY_INSTRUCTION() -> str:
-    return """[OBJECTIVE]
+    return f"""[OBJECTIVE]
 Extract Reimbursement Terms from a section of a payer-provider contract
 The output from this prompt will be used to populate downstream fields in the system.
 
@@ -536,6 +537,17 @@ Some types of Reimbursements are special and require additional thought.
 - Stop Loss and Outliers - These reimbursements usually consist of multiple rates and/or thresholds. When a Stop-Loss or Outlier is found, include the entire statement (which may be multiple sentences) as a single REIMB_TERM.
 - Discounts, Escalators and Premiums - These will also include multiple rates. Capture all relevant numbers
 
+[OUTPUT FORMAT]
+You MUST return your answer as a JSON list of dictionaries. Each dictionary must contain exactly two keys: "SERVICE_TERM" and "REIMB_TERM". Both values must be strings (not lists).
+
+Example output format:
+[
+  {{"SERVICE_TERM": "Service 1", "REIMB_TERM": "Reimb 1"}},
+  {{"SERVICE_TERM": "Service 2", "REIMB_TERM": "Reimb 2"}}
+]
+
+{JSON_LIST_OF_DICTS_FORMAT_INSTRUCTIONS}
+
 [ANALYSIS CONTEXT]"""
 
 
@@ -852,7 +864,7 @@ Populate a list of JSON dictionaries with the following fields:
 
 [OUTPUT FORMAT]
 The output should always be a JSON list/array of dictionaries, even when there is only one methodology.
-For any fields that don't apply to a particular methodology, use "N/A" as the value.
+{JSON_LIST_OF_DICTS_FORMAT_INSTRUCTIONS}
 
 [CONTEXT]"""
 
@@ -874,16 +886,20 @@ Populate a JSON dictionary with the following fields:
 {fields_text}
 
 [OUTPUT FORMAT]
-Briefly explain your reasoning, then write your JSON dictionary."""
+{JSON_DICT_FORMAT_INSTRUCTIONS}"""
 
 
 def FEE_SCHEDULE_BREAKOUT(
-    methodology, fee_schedule
+    methodology:str, fee_schedule:str
 ) -> Tuple[str, Callable[[str], dict]]:
     """Returns ONLY dynamic content for fee schedule breakout.
     Call FEE_SCHEDULE_BREAKOUT_INSTRUCTION() separately for the cached instruction.
     Note: Field definitions are now included in FEE_SCHEDULE_BREAKOUT_INSTRUCTION() for caching.
 
+    Args:
+        methodology: The reimbursement methodology string (already normalized to string in prompt_calls)
+        fee_schedule: The fee schedule string (already normalized to string in prompt_calls)
+
     Returns:
         Tuple of (prompt_string, parser_function) where parser expects JSON dict output.
     """
@@ -899,7 +915,6 @@ def GROUPER_BREAKOUT_INSTRUCTION() -> str:
     Contains objective, extraction rules, field definitions, and output format.
     Field definitions are loaded from investment_prompts.json.
     """
-    # Load fields from investment_prompts.json with resolved valid_values
     fields_text = _get_fields_text("grouper_breakout")
 
     return f"""[OBJECTIVE]
@@ -1151,6 +1166,10 @@ def CODE_EXPLICIT(service, methodology) -> Tuple[str, Callable[[str], dict]]:
     Call CODE_EXPLICIT_INSTRUCTION() separately for the cached instruction.
     Note: Field definitions are now included in CODE_EXPLICIT_INSTRUCTION() for caching.
 
+    Args:
+        service: The service term (already normalized to string upstream)
+        methodology: The reimbursement methodology (already normalized to string upstream)
+
     Returns:
         Tuple of (prompt_string, parser_function) where parser expects JSON dict output.
     """
@@ -1182,6 +1201,10 @@ Explain your answer in 1-2 sentences. Write your final answer in a properly-form
 def CODE_CATEGORY(service, choices) -> Tuple[str, Callable[[str], list]]:
     """Call CODE_CATEGORY_INSTRUCTION() separately for the cached instruction.
 
+    Args:
+        service: The service term (already normalized to string upstream)
+        choices: The valid descriptions to choose from
+
     Returns:
         Tuple of (prompt_string, parser_function) where parser expects JSON list output.
     """
@@ -1255,6 +1278,10 @@ Explain your answer, ensuring each point in the instructions is addressed. Then
 def CODE_IMPLICIT(service, choices) -> Tuple[str, Callable[[str], list]]:
     """Call CODE_IMPLICIT_INSTRUCTION() separately for the cached instruction.
 
+    Args:
+        service: The service term (already normalized to string upstream)
+        choices: The valid descriptions to choose from
+
     Returns:
         Tuple of (prompt_string, parser_function) where parser expects JSON list output.
     """
@@ -1325,6 +1352,12 @@ def FILL_BILL_TYPE(
 ) -> Tuple[str, Callable[[str], list]]:
     """Call FILL_BILL_TYPE_INSTRUCTION() separately for the cached instruction.
 
+    Args:
+        service: The service term (already normalized to string upstream)
+        choices: The valid bill type descriptions to choose from
+        reimb_term: The reimbursement term (already normalized to string upstream, optional)
+        exhibit_text: Additional exhibit context (optional)
+
     Returns:
         Tuple of (prompt_string, parser_function) where parser expects JSON list output.
     """
@@ -1346,6 +1379,7 @@ If the Service Term and Reimbursement Term together do not provide sufficient in
 {exhibit_text.replace('"', "'")}
 """
 
+    # service is already normalized to string upstream
     prompt = f"""[VALID DESCRIPTIONS]
 Here are the descriptions to choose from:
 {choices}

src/utils/crosswalk_utils.py

@@ -4,16 +4,16 @@ import logging
 import src.utils.string_utils as string_utils
 
 
-def apply_crosswalk(val: str, mapping: dict[str, str], default: str = "N/A") -> str:
+def apply_crosswalk(val: str | list, mapping: dict[str, str], default: str = "N/A") -> list:
     """Apply a crosswalk to a value
 
     Args:
-        val (str): Input value
+        val (str | list): Input value (can be string, list, or string representation of list)
         mapping (dict[str, str]): Mapping dictionary
         default (str | None, optional): Default value; if None and val is not found in the mapping, val is returned. Defaults to "".
 
     Returns:
-        str: Crosswalked value
+        list: Crosswalked value as a list
     """
     # Handle None or blank values defensively
     if val is None:

src/utils/json_utils.py

@@ -253,4 +253,4 @@ def parse_json_dict_or_list(raw_llm_output: str) -> dict[str, Any] | list[Any]:
     raise ValueError(
         f"No valid JSON object found in LLM output. "
         f"Output preview: {raw_llm_output[:200]}..."
-    )
+    )

src/utils/qa_qc_utils.py

@@ -1092,22 +1092,41 @@ def safe_token_to_str(tok: Any) -> str:
 
 def format_or_preserve_tins(val: Any) -> Tuple[str, bool]:
     """
-    Format TINs to 9 digits, preserving pipe-delimited multiple values.
+    Format TINs to 9 digits, handling JSON format (direct values or list of values).
 
     Args:
-        val: TIN value (may be pipe-delimited)
+        val: TIN value (may be a string, list of strings, or None/empty)
 
     Returns:
         Tuple of (formatted_tins, has_invalid)
     """
-    if pd.isna(val) or val is None or str(val).strip() == "":
-        return ("", True)
-
-    parts = [
-        part.strip()
-        for part in str(val).split("|")
-        if part is not None and part.strip() != ""
-    ]
+    # Handle list/array values (e.g., from PROV_GROUP_TIN, PROV_OTHER_TIN)
+    if isinstance(val, (list, tuple)) or (hasattr(val, '__iter__') and not isinstance(val, str)):
+        # Convert list to list of strings, filtering out empty values
+        if len(val) == 0:
+            return ("", True)
+        parts = [
+            str(item).strip()
+            for item in val
+            if item is not None and str(item).strip() != ""
+        ]
+        if not parts:
+            return ("", True)
+    else:
+        # Handle single value (string or None)
+        # Check for NaN/None/empty
+        try:
+            if pd.isna(val):
+                return ("", True)
+        except (ValueError, TypeError):
+            # pd.isna() can't handle arrays, but we've already handled lists above
+            pass
+        
+        if val is None or str(val).strip() == "":
+            return ("", True)
+        
+        # Single value - convert to list for consistent processing
+        parts = [str(val).strip()]
     out, any_invalid = [], False
 
     for part in parts: