更新

rag_eval/metrics/pipeline.py

@@ -94,6 +94,23 @@ class MetricPipeline:
                 reference=sample.ground_truth,
                 retrieved_contexts=sample.contexts,
             )
+        elif name == "noise_sensitivity":
+            coroutine = metric.ascore(
+                user_input=sample.question,
+                response=sample.answer,
+                reference=sample.ground_truth,
+                retrieved_contexts=sample.contexts,
+            )
+        elif name == "factual_correctness":
+            coroutine = metric.ascore(
+                response=sample.answer,
+                reference=sample.ground_truth,
+            )
+        elif name == "semantic_similarity":
+            coroutine = metric.ascore(
+                reference=sample.ground_truth,
+                response=sample.answer,
+            )
         else:
             raise ValueError(f"Unsupported metric: {name}")