diff --git a/mteb/benchmarks/benchmarks/benchmarks.py b/mteb/benchmarks/benchmarks/benchmarks.py
index a175a9de01..88fab8eb4f 100644
--- a/mteb/benchmarks/benchmarks/benchmarks.py
+++ b/mteb/benchmarks/benchmarks/benchmarks.py
@@ -2307,7 +2307,9 @@
             "RealMMRagTechSlidesRetrieval",
         ],
     ),
-    description="Realistic and multi-modal document retrieval benchmark.",
+    description="""REAL-MM-RAG is a realistic benchmark that reflects real-world multi-modal document retrieval challenges.
+                It includes infographic-rich documents such as slides, reports, and technical manuals with tables, charts, and figures, requiring models to integrate textual and visual evidence.
+                The benchmark features multi-modality, realistic queries, and accurate labeling for comprehensive evaluation.""",
     reference="https://arxiv.org/abs/2502.12342",
     citation=r"""
 @article{wasserman2025real,
diff --git a/mteb/leaderboard/benchmark_selector.py b/mteb/leaderboard/benchmark_selector.py
index bb7e3ee457..3395a9ecf9 100644
--- a/mteb/leaderboard/benchmark_selector.py
+++ b/mteb/leaderboard/benchmark_selector.py
@@ -46,7 +46,6 @@ class MenuEntry:
                         "MIEB(Img)",
                         "VisualDocumentRetrieval",
                         "JinaVDR",
-                        "REAL_MM_RAG"
                     ]
                 ),
             ),
diff --git a/mteb/tasks/Image/Any2AnyRetrieval/eng/RealMMRagBenchRetrieval.py b/mteb/tasks/Image/Any2AnyRetrieval/eng/RealMMRagBenchRetrieval.py
index f3c5e8eee1..3e571177d8 100644
--- a/mteb/tasks/Image/Any2AnyRetrieval/eng/RealMMRagBenchRetrieval.py
+++ b/mteb/tasks/Image/Any2AnyRetrieval/eng/RealMMRagBenchRetrieval.py
@@ -75,7 +75,9 @@ def _load_data(
 class RealMMRagFinReportRetrieval(AbsTaskAny2AnyRetrieval):
     metadata = TaskMetadata(
         name="RealMMRagFinReportRetrieval",
-        description="Retrieve associated pages according to questions.",
+        description="""Contains annual financial reports rich in text, tables, and figures from IBM’s public filings.
+                    Queries ask about financial results, trends, or statements across multiple years.
+                    Retrieval goal: find the specific report page containing the relevant financial information.""",
         reference="https://arxiv.org/abs/2502.12342",
         dataset={
             "path": "ibm-research/REAL-MM-RAG_FinReport_BEIR",
@@ -133,7 +135,9 @@ def load_data(self, **kwargs):
 class RealMMRagFinSlidesRetrieval(AbsTaskAny2AnyRetrieval):
     metadata = TaskMetadata(
         name="RealMMRagFinSlidesRetrieval",
-        description="Retrieve associated pages according to questions.",
+        description="""Comprises quarterly investor presentation slides combining tables, charts, and textual highlights.
+                    Queries focus on revenue trends, growth metrics, or business segments.
+                    Retrieval goal: retrieve the slide that presents the requested financial data or insight.""",
         reference="https://arxiv.org/abs/2502.12342",
         dataset={
             "path": "ibm-research/REAL-MM-RAG_FinSlides_BEIR",
@@ -190,7 +194,9 @@ def load_data(self, **kwargs):
 class RealMMRagTechReportRetrieval(AbsTaskAny2AnyRetrieval):
     metadata = TaskMetadata(
         name="RealMMRagTechReportRetrieval",
-        description="Retrieve associated pages according to questions.",
+        description="""Includes technical documentation and whitepapers on IBM storage and automation systems with text-heavy content and supporting visuals.
+                    Queries address specific technologies, architectures, or performance aspects.
+                    Retrieval goal: locate the report page providing the technical explanation or result.""",
         reference="https://arxiv.org/abs/2502.12342",
         dataset={
             "path": "ibm-research/REAL-MM-RAG_TechReport_BEIR",
@@ -247,7 +253,9 @@ def load_data(self, **kwargs):
 class RealMMRagTechSlidesRetrieval(AbsTaskAny2AnyRetrieval):
     metadata = TaskMetadata(
         name="RealMMRagTechSlidesRetrieval",
-        description="Retrieve associated pages according to questions.",
+        description="""Features technical presentation slides containing bullet points, flow diagrams, and schematic figures.
+                    Queries reflect realistic information-seeking about system design or AI and automation concepts.
+                    Retrieval goal: retrieve the slide that best answers the technical query through text and visuals.""",
         reference="https://arxiv.org/abs/2502.12342",
         dataset={
             "path": "ibm-research/REAL-MM-RAG_TechSlides_BEIR",