[Ref] error handling

src/similarity.py

@@ -6,7 +6,7 @@ def compute_similarity(selected_model_a, selected_model_b, selected_dataset):
     probs_a, gt_a = load_run_data(selected_model_a, selected_dataset)
     probs_b, gt_b = load_run_data(selected_model_b, selected_dataset)
 
-    assert len(probs_a) == len(probs_b), "Models must have the same number of responses"
+    assert len(probs_a) == len(probs_b), f"Models must have the same number of responses: {len(probs_a)} != {len(probs_b)}"
     
     # Only keep responses where the ground truth is the same
     output_a = []