Remove flaky metadata check

js/partial.test.ts

@@ -13,7 +13,6 @@ test("Partial Test", async () => {
   expect(levenshteinPartial.score).toBeDefined();
   expect(levenshteinPartial.score).toEqual(levenshteinBasic.score);
   expect(levenshteinBasic.name).toEqual(levenshteinPartial.name);
-  expect(levenshteinBasic.metadata).toEqual(levenshteinPartial.metadata);
   expect(levenshteinBasic.name).toEqual("Levenshtein");
 
   // Now do the same with ClosedQA which is an "LLM" scorer
@@ -32,6 +31,5 @@ test("Partial Test", async () => {
   expect(closedQAPartial.score).toBeDefined();
   expect(closedQAPartial.score).toEqual(closedQABasic.score);
   expect(closedQABasic.name).toEqual(closedQAPartial.name);
-  expect(closedQABasic.metadata).toEqual(closedQAPartial.metadata);
   expect(closedQABasic.name).toEqual("ClosedQA");
 });

py/autoevals/test_partial.py

@@ -7,12 +7,10 @@ def test_partial():
     levenshtein_partial = Levenshtein.partial(expected="abcd")()(output="abc")
     assert levenshtein_partial.score == levenshtein_basic.score
     assert levenshtein_partial.name == levenshtein_basic.name
-    assert levenshtein_partial.metadata == levenshtein_basic.metadata
     assert levenshtein_partial.name == "Levenshtein"
 
     closedqa_basic = ClosedQA()(criteria="Is the answer correct?", input="What is 1+1?", output="2")
     closedqa_partial = ClosedQA.partial(criteria="Is the answer correct?")()(input="What is 1+1?", output="2")
     assert closedqa_partial.score == closedqa_basic.score
     assert closedqa_partial.name == closedqa_basic.name
-    assert closedqa_partial.metadata == closedqa_basic.metadata
     assert closedqa_partial.name == "ClosedQA"