diff --git a/src/raglite/_search.py b/src/raglite/_search.py
index c65bfbc..1d43a2f 100644
--- a/src/raglite/_search.py
+++ b/src/raglite/_search.py
@@ -33,7 +33,7 @@ def search(query: str, *, config: RAGLiteConfig | None = None) -> tuple[list[Chu
 
 
 def vector_search(
-    query: str | FloatMatrix, *, config: RAGLiteConfig | None = None
+    query: str | FloatMatrix, *, oversample: int = 8, config: RAGLiteConfig | None = None
 ) -> tuple[list[ChunkId], list[float]]:
     """Search chunks using ANN vector search."""
     # Read the config.
@@ -66,7 +66,7 @@ def vector_search(
             results = session.exec(
                 select(ChunkEmbedding.chunk_id, distance)
                 .order_by(distance)
-                .limit(config.num_chunks)
+                .limit(config.num_chunks * oversample)
             )
             chunk_ids_, distance = zip(*results, strict=True)
             chunk_ids, similarity = np.asarray(chunk_ids_), 1.0 - np.asarray(distance)
@@ -79,7 +79,7 @@ def vector_search(
         from pynndescent import NNDescent
 
         multi_vector_indices, distance = cast(NNDescent, index).query(
-            query_embedding[np.newaxis, :], k=config.num_chunks
+            query_embedding[np.newaxis, :], k=config.num_chunks * oversample
         )
         similarity = 1 - distance[0, :]
         # Transform the multi-vector indices into chunk indices, and then to chunk ids.
@@ -175,7 +175,7 @@ def reciprocal_rank_fusion(
 
 
 def hybrid_search(
-    query: str, *, oversample: int = 4, config: RAGLiteConfig | None = None
+    query: str, *, oversample: int = 1, config: RAGLiteConfig | None = None
 ) -> tuple[list[ChunkId], list[float]]:
     """Search chunks by combining ANN vector search with BM25 keyword search."""
     config = config or RAGLiteConfig()