diff --git a/src/torchjd/aggregation/__init__.py b/src/torchjd/aggregation/__init__.py
index 8d1a9432..2bd845a2 100644
--- a/src/torchjd/aggregation/__init__.py
+++ b/src/torchjd/aggregation/__init__.py
@@ -1,8 +1,7 @@
 r"""
 When doing Jacobian descent, the Jacobian matrix has to be aggregated into a vector to store in the
 ``.grad`` fields of the model parameters. The
-The :class:`~torchjd.aggregation._aggregator_bases.Aggregator` is responsible for these
-aggregations.
+:class:`~torchjd.aggregation._aggregator_bases.Aggregator` is responsible for these aggregations.
 
 When using the :doc:`autogram <../autogram/index>` engine, we rather need to extract a vector
 of weights from the Gramian of the Jacobian. The