Fix TestGLM41VPPAccuracy test flakiness

python/sglang/test/run_eval.py

@@ -115,7 +115,11 @@ def run_eval(args):
         # VLM MMMU evaluation with fixed 100 examples by default
         from sglang.test.simple_eval_mmmu_vlm import MMMUVLMEval
 
-        eval_obj = MMMUVLMEval(args.num_examples, args.num_threads)
+        eval_obj = MMMUVLMEval(
+            args.num_examples,
+            args.num_threads,
+            response_answer_regex=getattr(args, "response_answer_regex", None),
+        )
     else:
         raise ValueError(f"Invalid eval name: {args.eval_name}")
 

python/sglang/test/simple_eval_mmmu_vlm.py

@@ -7,6 +7,7 @@
 
 import base64
 import io
+import re
 from typing import List, Optional, Tuple
 
 from datasets import concatenate_datasets, load_dataset
@@ -53,14 +54,22 @@ class MMMUVLMEval(Eval):
     }
 
     def __init__(
-        self, num_examples: Optional[int] = 100, num_threads: int = 32, seed: int = 42
+        self,
+        num_examples: Optional[int] = 100,
+        num_threads: int = 32,
+        seed: int = 42,
+        response_answer_regex: str = None,
     ):
         """Create MMMU VLM eval (Math subset, 100 fixed samples by default)."""
         self.num_examples = num_examples
         self.num_threads = num_threads
         self.seed = seed
         # Prepare samples deterministically across all MMMU subjects (validation split)
         self.samples = self._prepare_mmmu_samples(self.num_examples)
+        # For example, "<\|begin_of_box\|>foo<\|end_of_box\|>" could be used to extract "foo" as the answer from the response text
+        self.response_answer_regex = (
+            response_answer_regex if response_answer_regex is not None else "(.*)"
+        )
 
     @staticmethod
     def _to_data_uri(image: Image.Image) -> str:
@@ -205,6 +214,14 @@ def fn(sample: dict):
             # Sample
             response_text = sampler(prompt_messages)
             response_text = response_text or ""
+            match = (
+                re.search(self.response_answer_regex, response_text)
+                if response_text is not None
+                else None
+            )
+            response_text = (
+                match.group(1).strip() if match is not None else response_text
+            )
 
             # Parse and score
             gold = sample["answer"]

test/srt/test_pp_single_node.py

@@ -336,6 +336,8 @@ def setUpClass(cls):
                 "--chunked-prefill-size",
                 8192,
                 "--enable-multimodal",
+                "--reasoning-parser",
+                "glm45",
             ],
         )
 
@@ -350,10 +352,12 @@ def test_mmmu(self):
             eval_name="mmmu",
             num_examples=None,
             num_threads=32,
+            response_answer_regex="<\|begin_of_box\|>(.*)<\|end_of_box\|>",
         )
+
         metrics = run_eval(args)
         print(f"{metrics=}")
-        self.assertGreater(metrics["score"], 0.55)
+        self.assertGreater(metrics["score"], 0.45)
 
 
 if __name__ == "__main__":