diff --git a/README.md b/README.md
index 1822ff8..383fbe3 100644
--- a/README.md
+++ b/README.md
@@ -51,6 +51,7 @@ description: >
   Neural network architectures that operate directly on graph-structured data...
 topic_id: null          # OpenAlex topic delimiting the field
 seed_papers: []         # OpenAlex work IDs the crawl starts from
+exclude_works: []       # records whose metadata OpenAlex got wrong
 crawl:
   hop_depth: 2          # references, backwards from the seeds
   forward_depth: 1      # works citing the seeds, within topic_id
diff --git a/fields/gnn.yaml b/fields/gnn.yaml
index b995ff6..76acdba 100644
--- a/fields/gnn.yaml
+++ b/fields/gnn.yaml
@@ -31,6 +31,12 @@ seed_papers:
   - W2624431344  # Inductive Representation Learning on Large Graphs (2017)
   - W2766453196  # Graph Attention Networks (2017)
 
+# Works dropped at cleaning: OpenAlex records whose metadata is wrong in a way
+# the pipeline cannot detect, so they would surface in rankings under a name
+# that is not theirs.
+exclude_works:
+  - W2964015378  # cited 2347 times inside the graph, titled as a 2025 vulnerability-detection paper
+
 crawl:
   # How many citation hops to expand outward from the seeds.
   hop_depth: 2
diff --git a/src/graph/build.py b/src/graph/build.py
index 1922458..785d2cd 100644
--- a/src/graph/build.py
+++ b/src/graph/build.py
@@ -11,6 +11,7 @@
 import argparse
 import json
 import logging
+from collections.abc import Iterable
 from dataclasses import asdict, dataclass, field
 from datetime import date
 from pathlib import Path
@@ -44,6 +45,7 @@ class CleaningReport:
 
     raw_nodes: int = 0
     raw_edges: int = 0
+    excluded_nodes: int = 0
     duplicate_nodes: int = 0
     duplicate_edges: int = 0
     out_of_range_nodes: int = 0
@@ -104,6 +106,7 @@ def clean(
     edges: pd.DataFrame,
     *,
     date_range: DateRange | None = None,
+    exclude: Iterable[str] = (),
     largest_component_only: bool = True,
 ) -> tuple[pd.DataFrame, pd.DataFrame, CleaningReport]:
     """Reduce raw crawl tables to a consistent node set and edge set."""
@@ -112,6 +115,10 @@ def clean(
     nodes = nodes.drop_duplicates(subset="id")
     report.duplicate_nodes = report.raw_nodes - len(nodes)
 
+    excluded = nodes["id"].isin(set(exclude))
+    report.excluded_nodes = int(excluded.sum())
+    nodes = nodes[~excluded]
+
     if date_range is not None:
         in_range = nodes["publication_year"].map(date_range.contains)
         report.out_of_range_nodes = int((~in_range).sum())
@@ -266,6 +273,7 @@ def build_field(
         nodes,
         edges,
         date_range=config.crawl.date_range,
+        exclude=config.exclude_works,
         largest_component_only=largest_component_only,
     )
     surviving = set(nodes["id"])
diff --git a/src/graph/config.py b/src/graph/config.py
index 89219fe..cbf5ab1 100644
--- a/src/graph/config.py
+++ b/src/graph/config.py
@@ -32,6 +32,7 @@ class FieldConfig(BaseModel):
     description: str
     topic_id: str | None = None
     seed_papers: list[str] = Field(default_factory=list)
+    exclude_works: list[str] = Field(default_factory=list)
     crawl: CrawlConfig
 
     @field_validator("description")
diff --git a/tests/test_build.py b/tests/test_build.py
index 0855e18..36d56f9 100644
--- a/tests/test_build.py
+++ b/tests/test_build.py
@@ -62,6 +62,16 @@ def test_edges_past_the_crawl_horizon_are_dropped():
     assert list(edges["target"]) == ["B"]
 
 
+def test_excluded_works_are_dropped_with_their_edges():
+    nodes, edges = frames([node("A"), node("B"), node("C")], [("A", "B"), ("A", "C")])
+
+    nodes, edges, report = clean(nodes, edges, exclude=["B"])
+
+    assert report.excluded_nodes == 1
+    assert sorted(nodes["id"]) == ["A", "C"]
+    assert list(edges.itertuples(index=False, name=None)) == [("A", "C")]
+
+
 def test_nodes_outside_the_date_range_are_dropped_with_their_edges():
     nodes, edges = frames(
         [node("A", year=2015), node("B", year=1970), node("C", year=2016)],