diff --git a/js/llm.ts b/js/llm.ts
index 46822b3..f37c5da 100644
--- a/js/llm.ts
+++ b/js/llm.ts
@@ -125,8 +125,6 @@ export async function OpenAIClassifier<RenderArgs, Output>(
     content: m.content && mustache.render(m.content as string, renderArgs),
   }));
 
-  let ret = null;
-  let validityScore = 1;
   try {
     const resp = await cachedChatCompletion(
       {
@@ -145,7 +143,7 @@ export async function OpenAIClassifier<RenderArgs, Output>(
     );
 
     if (resp.choices.length > 0) {
-      ret = {
+      return {
         name,
         ...parseResponse(resp.choices[0].message!, choiceScores),
       };
@@ -153,15 +151,12 @@ export async function OpenAIClassifier<RenderArgs, Output>(
       throw new Error("Empty response from OpenAI");
     }
   } catch (error) {
-    validityScore = 0;
-    ret = {
+    return {
       name,
       score: 0,
       error: `${error}`,
     };
   }
-
-  return ret;
 }
 
 function parseResponse(
diff --git a/py/autoevals/llm.py b/py/autoevals/llm.py
index 6fa21ff..2ac8363 100644
--- a/py/autoevals/llm.py
+++ b/py/autoevals/llm.py
@@ -153,26 +153,18 @@ def _postprocess_response(self, resp):
             raise ValueError("Empty response from OpenAI")
 
     async def _run_eval_async(self, output, expected, **kwargs):
-        validity_score = 1
         try:
             return self._postprocess_response(
                 await arun_cached_request(**self._request_args(output, expected, **kwargs))
             )
         except Exception as e:
-            validity_score = 0
             return Score(name=self.name, score=0, error=e)
-        finally:
-            current_span().log(scores={f"{self._name()} parsed": validity_score})
 
     def _run_eval_sync(self, output, expected, **kwargs):
-        validity_score = 1
         try:
             return self._postprocess_response(run_cached_request(**self._request_args(output, expected, **kwargs)))
         except Exception as e:
-            validity_score = 0
             return Score(name=self.name, score=0, error=e)
-        finally:
-            current_span().log(scores={f"{self._name()} parsed": validity_score})
 
 
 @dataclass