refactor: split BRIGHT benchmark into individual subset tasks

Author: whybe-choi

mteb/benchmarks/benchmarks/benchmarks.py

@@ -1165,7 +1165,22 @@
 BRIGHT = Benchmark(
     name="BRIGHT",
     display_name="Reasoning Retrieval",
-    tasks=get_tasks(tasks=["BrightRetrieval"], eval_splits=["standard"]),
+    tasks=get_tasks(
+        tasks=[
+            "BrightBiologyRetrieval",
+            "BrightEarthScienceRetrieval",
+            "BrightEconomicsRetrieval",
+            "BrightPsychologyRetrieval",
+            "BrightRoboticsRetrieval",
+            "BrightStackoverflowRetrieval",
+            "BrightSustainableLivingRetrieval",
+            "BrightPonyRetrieval",
+            "BrightLeetcodeRetrieval",
+            "BrightAopsRetrieval",
+            "BrightTheoremQATheoremsRetrieval",
+            "BrightTheoremQAQuestionsRetrieval",
+        ],
+    ),
     description="""BRIGHT: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrieval.
     BRIGHT is the first text retrieval
     benchmark that requires intensive reasoning to retrieve relevant documents with
@@ -1186,12 +1201,17 @@
 
 BRIGHT_LONG = Benchmark(
     name="BRIGHT (long)",
-    tasks=MTEBTasks(
-        (
-            get_task(
-                "BrightLongRetrieval",
-            ),
-        )
+    tasks=get_tasks(
+        tasks=[
+            "BrightBiologyLongRetrieval",
+            "BrightEarthScienceLongRetrieval",
+            "BrightEconomicsLongRetrieval",
+            "BrightPsychologyLongRetrieval",
+            "BrightRoboticsLongRetrieval",
+            "BrightStackoverflowLongRetrieval",
+            "BrightSustainableLivingLongRetrieval",
+            "BrightPonyLongRetrieval",
+        ],
     ),
     description="""BRIGHT: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrieval.
 BRIGHT is the first text retrieval
@@ -1603,8 +1623,7 @@
             "TRECCOVID-NL",
         ],
     ),
-    description="BEIR-NL is a Dutch adaptation of the publicly available BEIR benchmark, created through automated "
-    "translation.",
+    description="BEIR-NL is a Dutch adaptation of the publicly available BEIR benchmark, created through automated translation.",
     reference="https://arxiv.org/abs/2412.08329",
     contacts=["nikolay-banar"],
     citation=r"""

mteb/tasks/Retrieval/__init__.py

@@ -42,6 +42,7 @@
 from .eng.BarExamQARetrieval import *
 from .eng.BillSumCARetrieval import *
 from .eng.BillSumUSRetrieval import *
+from .eng.BrightLongRetrieval import *
 from .eng.BrightRetrieval import *
 from .eng.BuiltBenchRetrieval import *
 from .eng.ChatDoctorRetrieval import *