always print benchmark question 2 and 3

backend/routes/questions.py

@@ -35,9 +35,11 @@ async def get_benchmark_questions(session_id: str):
             # Essayer de charger les questions single-shot directement avec le nom de config
             single_dataset = load_dataset(dataset_repo_id, 'single_shot_questions')
             if single_dataset and len(single_dataset['train']) > 0:
-                # Get a random sample (up to 2) from single-shot questions
-                sample_indices = random.sample(range(len(single_dataset['train'])), min(2, len(single_dataset['train'])))
-                for idx in sample_indices:
+                # Prendre 2 questions à partir de l'index 1 (en évitant la première question)
+                start_idx = 1
+                max_questions = min(2, max(0, len(single_dataset['train']) - start_idx))
+                for i in range(max_questions):
+                    idx = start_idx + i
                     questions.append({
                         "id": str(idx),
                         "question": single_dataset['train'][idx].get("question", ""),
@@ -53,10 +55,12 @@ async def get_benchmark_questions(session_id: str):
             if len(questions) < 2:
                 multi_dataset = load_dataset(dataset_repo_id, 'multi_hop_questions')
                 if multi_dataset and len(multi_dataset['train']) > 0:
-                    # Get remaining questions from multi-hop questions
+                    # Prendre les questions multi-hop pour compléter, en évitant aussi la première
+                    start_idx = 1
                     remaining = 2 - len(questions)
-                    sample_indices = random.sample(range(len(multi_dataset['train'])), min(remaining, len(multi_dataset['train'])))
-                    for idx in sample_indices:
+                    max_questions = min(remaining, max(0, len(multi_dataset['train']) - start_idx))
+                    for i in range(max_questions):
+                        idx = start_idx + i
                         questions.append({
                             "id": str(idx),
                             "question": multi_dataset['train'][idx].get("question", ""),