deepeval 3.7.4__py3-none-any.whl → 3.7.5__py3-none-any.whl

deepeval/metrics/multimodal_metrics/image_coherence/image_coherence.py

@@ -2,7 +2,7 @@ import asyncio
 from typing import Optional, List, Tuple, Union
 
 from deepeval.metrics import BaseMultimodalMetric
-from deepeval.test_case import MLLMTestCaseParams, MLLMTestCase, MLLMImage
+from deepeval.test_case import LLMTestCaseParams, LLMTestCase, MLLMImage
 from deepeval.metrics.multimodal_metrics.image_coherence.template import (
     ImageCoherenceTemplate,
 )
@@ -10,32 +10,35 @@ from deepeval.metrics.utils import (
     construct_verbose_logs,
     trimAndLoadJson,
     check_mllm_test_case_params,
-    initialize_multimodal_model,
+    initialize_model,
 )
-from deepeval.models import DeepEvalBaseMLLM
+from deepeval.models import DeepEvalBaseLLM
 from deepeval.metrics.multimodal_metrics.image_coherence.schema import (
     ReasonScore,
 )
 from deepeval.metrics.indicator import metric_progress_indicator
-from deepeval.utils import get_or_create_event_loop
+from deepeval.utils import (
+    get_or_create_event_loop,
+    convert_to_multi_modal_array,
+)
 
 
 class ImageCoherenceMetric(BaseMultimodalMetric):
-    _required_params: List[MLLMTestCaseParams] = [
-        MLLMTestCaseParams.INPUT,
-        MLLMTestCaseParams.ACTUAL_OUTPUT,
+    _required_params: List[LLMTestCaseParams] = [
+        LLMTestCaseParams.INPUT,
+        LLMTestCaseParams.ACTUAL_OUTPUT,
     ]
 
     def __init__(
         self,
-        model: Optional[Union[str, DeepEvalBaseMLLM]] = None,
+        model: Optional[Union[str, DeepEvalBaseLLM]] = None,
         threshold: float = 0.5,
         async_mode: bool = True,
         strict_mode: bool = False,
         verbose_mode: bool = False,
         max_context_size: Optional[int] = None,
     ):
-        self.model, self.using_native_model = initialize_multimodal_model(model)
+        self.model, self.using_native_model = initialize_model(model)
         self.evaluation_model = self.model.get_model_name()
         self.threshold = 1 if strict_mode else threshold
         self.strict_mode = strict_mode
@@ -45,13 +48,13 @@ class ImageCoherenceMetric(BaseMultimodalMetric):
 
     def measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, None, None, self
+            test_case, self._required_params, None, None, self, self.model
         )
         self.evaluation_cost = 0 if self.using_native_model else None
         with metric_progress_indicator(
@@ -68,7 +71,9 @@ class ImageCoherenceMetric(BaseMultimodalMetric):
                     )
                 )
             else:
-                actual_output = test_case.actual_output
+                actual_output = convert_to_multi_modal_array(
+                    test_case.actual_output
+                )
                 self.contexts_above = []
                 self.contexts_below = []
                 self.scores = []
@@ -145,13 +150,13 @@ class ImageCoherenceMetric(BaseMultimodalMetric):
 
     async def a_measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, None, None, self
+            test_case, self._required_params, None, None, self, self.model
         )
         self.evaluation_cost = 0 if self.using_native_model else None
         with metric_progress_indicator(
@@ -160,7 +165,9 @@ class ImageCoherenceMetric(BaseMultimodalMetric):
             _show_indicator=_show_indicator,
             _in_component=_in_component,
         ):
-            actual_output = test_case.actual_output
+            actual_output = convert_to_multi_modal_array(
+                test_case.actual_output
+            )
             self.contexts_above = []
             self.contexts_below = []
             self.scores = []
@@ -253,7 +260,7 @@ class ImageCoherenceMetric(BaseMultimodalMetric):
         instructions = ImageCoherenceTemplate.evaluate_image_coherence(
             context_above, context_below
         )
-        prompt = [instructions] + [image]
+        prompt = f"{instructions} \nImages: {image}"
         if self.using_native_model:
             res, cost = self.model.generate(prompt, ReasonScore)
             self.evaluation_cost += cost
@@ -278,7 +285,7 @@ class ImageCoherenceMetric(BaseMultimodalMetric):
         instructions = ImageCoherenceTemplate.evaluate_image_coherence(
             context_above, context_below
         )
-        prompt = [instructions] + [image]
+        prompt = f"{instructions} \nImages: {image}"
         if self.using_native_model:
             res, cost = await self.model.a_generate(prompt, schema=ReasonScore)
             self.evaluation_cost += cost

deepeval/metrics/multimodal_metrics/image_editing/image_editing.py

@@ -4,38 +4,41 @@ import math
 import textwrap
 
 from deepeval.metrics import BaseMultimodalMetric
-from deepeval.test_case import MLLMTestCaseParams, MLLMTestCase, MLLMImage
+from deepeval.test_case import LLMTestCaseParams, LLMTestCase, MLLMImage
 from deepeval.metrics.multimodal_metrics.image_editing.template import (
     ImageEditingTemplate,
 )
-from deepeval.utils import get_or_create_event_loop
+from deepeval.utils import (
+    get_or_create_event_loop,
+    convert_to_multi_modal_array,
+)
 from deepeval.metrics.utils import (
     construct_verbose_logs,
     trimAndLoadJson,
     check_mllm_test_case_params,
-    initialize_multimodal_model,
+    initialize_model,
 )
-from deepeval.models import DeepEvalBaseMLLM
+from deepeval.models import DeepEvalBaseLLM
 from deepeval.metrics.multimodal_metrics.image_editing.schema import ReasonScore
 from deepeval.metrics.indicator import metric_progress_indicator
 
 
 class ImageEditingMetric(BaseMultimodalMetric):
 
-    _required_params: List[MLLMTestCaseParams] = [
-        MLLMTestCaseParams.INPUT,
-        MLLMTestCaseParams.ACTUAL_OUTPUT,
+    _required_params: List[LLMTestCaseParams] = [
+        LLMTestCaseParams.INPUT,
+        LLMTestCaseParams.ACTUAL_OUTPUT,
     ]
 
     def __init__(
         self,
-        model: Optional[Union[str, DeepEvalBaseMLLM]] = None,
+        model: Optional[Union[str, DeepEvalBaseLLM]] = None,
         threshold: float = 0.5,
         async_mode: bool = True,
         strict_mode: bool = False,
         verbose_mode: bool = False,
     ):
-        self.model, self.using_native_model = initialize_multimodal_model(model)
+        self.model, self.using_native_model = initialize_model(model)
         self.evaluation_model = self.model.get_model_name()
         self.threshold = 1 if strict_mode else threshold
         self.strict_mode = strict_mode
@@ -44,13 +47,13 @@ class ImageEditingMetric(BaseMultimodalMetric):
 
     def measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, 1, 1, self
+            test_case, self._required_params, 1, 1, self, self.model
         )
 
         self.evaluation_cost = 0 if self.using_native_model else None
@@ -68,12 +71,14 @@ class ImageEditingMetric(BaseMultimodalMetric):
                     )
                 )
             else:
-                input_texts, input_images = self.separate_images_from_text(
-                    test_case.input
-                )
-                _, output_images = self.separate_images_from_text(
+                input = convert_to_multi_modal_array(test_case.input)
+                actual_output = convert_to_multi_modal_array(
                     test_case.actual_output
                 )
+                input_texts, input_images = self.separate_images_from_text(
+                    input
+                )
+                _, output_images = self.separate_images_from_text(actual_output)
 
                 self.SC_scores, self.SC_reasoning = (
                     self._evaluate_semantic_consistency(
@@ -107,13 +112,13 @@ class ImageEditingMetric(BaseMultimodalMetric):
 
     async def a_measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, 1, 1, self
+            test_case, self._required_params, 1, 1, self, self.model
         )
 
         self.evaluation_cost = 0 if self.using_native_model else None
@@ -123,12 +128,12 @@ class ImageEditingMetric(BaseMultimodalMetric):
             _show_indicator=_show_indicator,
             _in_component=_in_component,
         ):
-            input_texts, input_images = self.separate_images_from_text(
-                test_case.input
-            )
-            _, output_images = self.separate_images_from_text(
+            input = convert_to_multi_modal_array(test_case.input)
+            actual_output = convert_to_multi_modal_array(
                 test_case.actual_output
             )
+            input_texts, input_images = self.separate_images_from_text(input)
+            _, output_images = self.separate_images_from_text(actual_output)
             (self.SC_scores, self.SC_reasoning), (
                 self.PQ_scores,
                 self.PQ_reasoning,

deepeval/metrics/multimodal_metrics/image_helpfulness/image_helpfulness.py

@@ -2,7 +2,7 @@ import asyncio
 from typing import Optional, List, Tuple, Union
 
 from deepeval.metrics import BaseMultimodalMetric
-from deepeval.test_case import MLLMTestCaseParams, MLLMTestCase, MLLMImage
+from deepeval.test_case import LLMTestCaseParams, LLMTestCase, MLLMImage
 from deepeval.metrics.multimodal_metrics.image_helpfulness.template import (
     ImageHelpfulnessTemplate,
 )
@@ -10,33 +10,36 @@ from deepeval.metrics.utils import (
     construct_verbose_logs,
     trimAndLoadJson,
     check_mllm_test_case_params,
-    initialize_multimodal_model,
+    initialize_model,
 )
-from deepeval.models import DeepEvalBaseMLLM
+from deepeval.models import DeepEvalBaseLLM
 from deepeval.metrics.multimodal_metrics.image_helpfulness.schema import (
     ReasonScore,
 )
 from deepeval.metrics.indicator import metric_progress_indicator
-from deepeval.utils import get_or_create_event_loop
+from deepeval.utils import (
+    get_or_create_event_loop,
+    convert_to_multi_modal_array,
+)
 
 
 class ImageHelpfulnessMetric(BaseMultimodalMetric):
 
-    _required_params: List[MLLMTestCaseParams] = [
-        MLLMTestCaseParams.INPUT,
-        MLLMTestCaseParams.ACTUAL_OUTPUT,
+    _required_params: List[LLMTestCaseParams] = [
+        LLMTestCaseParams.INPUT,
+        LLMTestCaseParams.ACTUAL_OUTPUT,
     ]
 
     def __init__(
         self,
-        model: Optional[Union[str, DeepEvalBaseMLLM]] = None,
+        model: Optional[Union[str, DeepEvalBaseLLM]] = None,
         threshold: float = 0.5,
         async_mode: bool = True,
         strict_mode: bool = False,
         verbose_mode: bool = False,
         max_context_size: Optional[int] = None,
     ):
-        self.model, self.using_native_model = initialize_multimodal_model(model)
+        self.model, self.using_native_model = initialize_model(model)
         self.evaluation_model = self.model.get_model_name()
         self.threshold = 1 if strict_mode else threshold
         self.strict_mode = strict_mode
@@ -46,13 +49,13 @@ class ImageHelpfulnessMetric(BaseMultimodalMetric):
 
     def measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, None, None, self
+            test_case, self._required_params, None, None, self, self.model
         )
         self.evaluation_cost = 0 if self.using_native_model else None
         with metric_progress_indicator(
@@ -69,7 +72,9 @@ class ImageHelpfulnessMetric(BaseMultimodalMetric):
                     )
                 )
             else:
-                actual_output = test_case.actual_output
+                actual_output = convert_to_multi_modal_array(
+                    test_case.actual_output
+                )
                 self.contexts_above = []
                 self.contexts_below = []
                 self.scores = []
@@ -146,13 +151,13 @@ class ImageHelpfulnessMetric(BaseMultimodalMetric):
 
     async def a_measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, None, None, self
+            test_case, self._required_params, None, None, self, self.model
         )
         self.evaluation_cost = 0 if self.using_native_model else None
         with metric_progress_indicator(
@@ -161,7 +166,9 @@ class ImageHelpfulnessMetric(BaseMultimodalMetric):
             _show_indicator=_show_indicator,
             _in_component=_in_component,
         ):
-            actual_output = test_case.actual_output
+            actual_output = convert_to_multi_modal_array(
+                test_case.actual_output
+            )
             self.contexts_above = []
             self.contexts_below = []
             self.scores = []
@@ -254,7 +261,7 @@ class ImageHelpfulnessMetric(BaseMultimodalMetric):
         instructions = ImageHelpfulnessTemplate.evaluate_image_helpfulness(
             context_above, context_below
         )
-        prompt = [instructions] + [image]
+        prompt = f"{instructions} \nImages: {image}"
         if self.using_native_model:
             res, cost = self.model.generate(prompt, schema=ReasonScore)
             self.evaluation_cost += cost
@@ -279,7 +286,7 @@ class ImageHelpfulnessMetric(BaseMultimodalMetric):
         instructions = ImageHelpfulnessTemplate.evaluate_image_helpfulness(
             context_above, context_below
         )
-        prompt = [instructions] + [image]
+        prompt = f"{instructions} \nImages: {image}"
         if self.using_native_model:
             res, cost = await self.model.a_generate(prompt, schema=ReasonScore)
             self.evaluation_cost += cost

deepeval/metrics/multimodal_metrics/image_reference/image_reference.py

@@ -2,7 +2,7 @@ import asyncio
 from typing import Optional, List, Tuple, Union
 
 from deepeval.metrics import BaseMultimodalMetric
-from deepeval.test_case import MLLMTestCaseParams, MLLMTestCase, MLLMImage
+from deepeval.test_case import LLMTestCaseParams, LLMTestCase, MLLMImage
 from deepeval.metrics.multimodal_metrics.image_reference.template import (
     ImageReferenceTemplate,
 )
@@ -10,33 +10,36 @@ from deepeval.metrics.utils import (
     construct_verbose_logs,
     trimAndLoadJson,
     check_mllm_test_case_params,
-    initialize_multimodal_model,
+    initialize_model,
 )
-from deepeval.models import DeepEvalBaseMLLM
+from deepeval.models import DeepEvalBaseLLM
 from deepeval.metrics.multimodal_metrics.image_reference.schema import (
     ReasonScore,
 )
 from deepeval.metrics.indicator import metric_progress_indicator
-from deepeval.utils import get_or_create_event_loop
+from deepeval.utils import (
+    get_or_create_event_loop,
+    convert_to_multi_modal_array,
+)
 
 
 class ImageReferenceMetric(BaseMultimodalMetric):
 
-    _required_params: List[MLLMTestCaseParams] = [
-        MLLMTestCaseParams.INPUT,
-        MLLMTestCaseParams.ACTUAL_OUTPUT,
+    _required_params: List[LLMTestCaseParams] = [
+        LLMTestCaseParams.INPUT,
+        LLMTestCaseParams.ACTUAL_OUTPUT,
     ]
 
     def __init__(
         self,
-        model: Optional[Union[str, DeepEvalBaseMLLM]] = None,
+        model: Optional[Union[str, DeepEvalBaseLLM]] = None,
         threshold: float = 0.5,
         async_mode: bool = True,
         strict_mode: bool = False,
         verbose_mode: bool = False,
         max_context_size: Optional[int] = None,
     ):
-        self.model, self.using_native_model = initialize_multimodal_model(model)
+        self.model, self.using_native_model = initialize_model(model)
         self.evaluation_model = self.model.get_model_name()
         self.threshold = 1 if strict_mode else threshold
         self.strict_mode = strict_mode
@@ -46,13 +49,13 @@ class ImageReferenceMetric(BaseMultimodalMetric):
 
     def measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, None, None, self
+            test_case, self._required_params, None, None, self, self.model
         )
         self.evaluation_cost = 0 if self.using_native_model else None
         with metric_progress_indicator(
@@ -69,7 +72,9 @@ class ImageReferenceMetric(BaseMultimodalMetric):
                     )
                 )
             else:
-                actual_output = test_case.actual_output
+                actual_output = convert_to_multi_modal_array(
+                    test_case.actual_output
+                )
                 self.contexts_above = []
                 self.contexts_below = []
                 self.scores = []
@@ -146,13 +151,13 @@ class ImageReferenceMetric(BaseMultimodalMetric):
 
     async def a_measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
     ) -> float:
         check_mllm_test_case_params(
-            test_case, self._required_params, None, None, self
+            test_case, self._required_params, None, None, self, self.model
         )
         self.evaluation_cost = 0 if self.using_native_model else None
         with metric_progress_indicator(
@@ -161,7 +166,9 @@ class ImageReferenceMetric(BaseMultimodalMetric):
             _show_indicator=_show_indicator,
             _in_component=_in_component,
         ):
-            actual_output = test_case.actual_output
+            actual_output = convert_to_multi_modal_array(
+                test_case.actual_output
+            )
             self.contexts_above = []
             self.contexts_below = []
             self.scores = []
@@ -254,7 +261,7 @@ class ImageReferenceMetric(BaseMultimodalMetric):
         instructions = ImageReferenceTemplate.evaluate_image_reference(
             context_above, context_below
         )
-        prompt = [instructions] + [image]
+        prompt = f"{instructions} \nImages: {image}"
         if self.using_native_model:
             res, cost = self.model.generate(prompt, schema=ReasonScore)
             self.evaluation_cost += cost
@@ -279,7 +286,7 @@ class ImageReferenceMetric(BaseMultimodalMetric):
         instructions = ImageReferenceTemplate.evaluate_image_reference(
             context_above, context_below
         )
-        prompt = [instructions] + [image]
+        prompt = f"{instructions} \nImages: {image}"
         if self.using_native_model:
             res, cost = await self.model.a_generate(prompt, schema=ReasonScore)
             self.evaluation_cost += cost

deepeval/metrics/multimodal_metrics/multimodal_g_eval/multimodal_g_eval.py

@@ -1,11 +1,11 @@
 """LLM evaluated metric based on the GEval framework: https://arxiv.org/pdf/2303.16634.pdf"""
 
 from typing import Optional, List, Tuple, Type, Union
-from deepeval.models import DeepEvalBaseMLLM
+from deepeval.models import DeepEvalBaseLLM
 from deepeval.metrics import BaseMultimodalMetric
 from deepeval.test_case import (
-    MLLMTestCaseParams,
-    MLLMTestCase,
+    LLMTestCaseParams,
+    LLMTestCase,
 )
 from deepeval.metrics.multimodal_metrics.multimodal_g_eval.template import (
     MultimodalGEvalTemplate,
@@ -17,7 +17,7 @@ from deepeval.metrics.multimodal_metrics.multimodal_g_eval.schema import (
 from deepeval.utils import get_or_create_event_loop, prettify_list
 from deepeval.metrics.indicator import metric_progress_indicator
 from deepeval.metrics.utils import (
-    initialize_multimodal_model,
+    initialize_model,
     check_mllm_test_case_params,
     construct_verbose_logs,
     trimAndLoadJson,
@@ -42,11 +42,11 @@ class MultimodalGEval(BaseMultimodalMetric):
     def __init__(
         self,
         name: str,
-        evaluation_params: List[MLLMTestCaseParams],
+        evaluation_params: List[LLMTestCaseParams],
         criteria: Optional[str] = None,
         evaluation_steps: Optional[List[str]] = None,
         rubric: Optional[List[Rubric]] = None,
-        model: Optional[Union[str, DeepEvalBaseMLLM]] = None,
+        model: Optional[Union[str, DeepEvalBaseLLM]] = None,
         threshold: float = 0.5,
         top_logprobs: int = 20,
         async_mode: bool = True,
@@ -62,7 +62,7 @@ class MultimodalGEval(BaseMultimodalMetric):
         self.evaluation_params = evaluation_params
         self.criteria = criteria
         self.rubric = validate_and_sort_rubrics(rubric)
-        self.model, self.using_native_model = initialize_multimodal_model(model)
+        self.model, self.using_native_model = initialize_model(model)
         self.evaluation_model = self.model.get_model_name()
         self.evaluation_steps = (
             evaluation_steps
@@ -79,7 +79,7 @@ class MultimodalGEval(BaseMultimodalMetric):
 
     def measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _log_metric_to_confident: bool = True,
@@ -87,7 +87,7 @@ class MultimodalGEval(BaseMultimodalMetric):
     ) -> float:
 
         check_mllm_test_case_params(
-            test_case, self.evaluation_params, None, None, self
+            test_case, self.evaluation_params, None, None, self, self.model
         )
 
         self.evaluation_cost = 0 if self.using_native_model else None
@@ -134,7 +134,7 @@ class MultimodalGEval(BaseMultimodalMetric):
 
     async def a_measure(
         self,
-        test_case: MLLMTestCase,
+        test_case: LLMTestCase,
         _show_indicator: bool = True,
         _in_component: bool = False,
         _additional_context: Optional[str] = None,
@@ -142,7 +142,7 @@ class MultimodalGEval(BaseMultimodalMetric):
     ) -> float:
 
         check_mllm_test_case_params(
-            test_case, self.evaluation_params, None, None, self
+            test_case, self.evaluation_params, None, None, self, self.model
         )
 
         self.evaluation_cost = 0 if self.using_native_model else None
@@ -185,15 +185,15 @@ class MultimodalGEval(BaseMultimodalMetric):
             criteria=self.criteria, parameters=g_eval_params_str
         )
         if self.using_native_model:
-            res, cost = await self.model.a_generate([prompt], schema=Steps)
+            res, cost = await self.model.a_generate(prompt, schema=Steps)
             self.evaluation_cost += cost
             return res.steps
         else:
             try:
-                res: Steps = await self.model.a_generate([prompt], schema=Steps)
+                res: Steps = await self.model.a_generate(prompt, schema=Steps)
                 return res.steps
             except TypeError:
-                res = await self.model.a_generate([prompt])
+                res = await self.model.a_generate(prompt)
                 data = trimAndLoadJson(res, self)
                 return data["steps"]
 
@@ -208,20 +208,20 @@ class MultimodalGEval(BaseMultimodalMetric):
             criteria=self.criteria, parameters=g_eval_params_str
         )
         if self.using_native_model:
-            res, cost = self.model.generate([prompt], schema=Steps)
+            res, cost = self.model.generate(prompt, schema=Steps)
             self.evaluation_cost += cost
             return res.steps
         else:
             try:
-                res: Steps = self.model.generate([prompt], schema=Steps)
+                res: Steps = self.model.generate(prompt, schema=Steps)
                 return res.steps
             except TypeError:
-                res = self.model.generate([prompt])
+                res = self.model.generate(prompt)
                 data = trimAndLoadJson(res, self)
                 return data["steps"]
 
     async def _a_evaluate(
-        self, test_case: MLLMTestCase, _additional_context: Optional[str] = None
+        self, test_case: LLMTestCase, _additional_context: Optional[str] = None
     ) -> Tuple[Union[int, float], str]:
         test_case_list = construct_test_case_list(
             self.evaluation_params, test_case
@@ -296,7 +296,7 @@ class MultimodalGEval(BaseMultimodalMetric):
                     return data["score"], data["reason"]
 
     def _evaluate(
-        self, test_case: MLLMTestCase, _additional_context: Optional[str] = None
+        self, test_case: LLMTestCase, _additional_context: Optional[str] = None
     ) -> Tuple[Union[int, float], str]:
         test_case_list = construct_test_case_list(
             self.evaluation_params, test_case