Update README

README.md

@@ -144,7 +144,7 @@ from openai import OpenAI
 
 client = OpenAI()  # or any OpenAI-compatible client
 
-def evaluate(question, response, rubrics, judge_model="gpt-4o-mini"):
+def evaluate(question, response, rubrics, judge_model="openai/gpt-5.4"):
     """Judge all rubrics in one call, return weighted score."""
     rubrics_text = "\n\n".join(
         f"**Rubric {r['rubric_number']}** (weight {r['rubric_weight']:+d})\n{r['rubric_detail']}"
@@ -174,7 +174,7 @@ def evaluate(question, response, rubrics, judge_model="gpt-4o-mini"):
 ds = load_dataset("humanlaya-data-lab/OneMillion-Bench", "natural_science", split="test")
 for entry in ds.select(range(3)):
     response = client.chat.completions.create(
-        model="gpt-4o-mini",
+        model="openai/gpt-5.4",
         messages=[{"role": "user", "content": entry["question"]}],
     ).choices[0].message.content
     result = evaluate(entry["question"], response, entry["rubrics"])
@@ -183,4 +183,4 @@ for entry in ds.select(range(3)):
 
 ## License
 
-Apache 2.0
+Apache 2.0