diff --git a/faiss_vector_posting.go b/faiss_vector_posting.go
index bf35f99..338e05d 100644
--- a/faiss_vector_posting.go
+++ b/faiss_vector_posting.go
@@ -409,14 +409,9 @@ func (sb *SegmentBase) UpdateFieldStats(stats segment.FieldStats) {
 			continue
 		}
 
-		_, _, numVecs, _, minDocID, maxDocID, indexBytesLen, _ := getVectorSectionContentOffsets(sb, pos)
+		_, _, numVecs, _, indexBytesLen, _ := getVectorSectionContentOffsets(sb, pos)
 
 		stats.Store("num_vectors", fieldName, numVecs)
 		stats.Store("vector_index_bytes", fieldName, indexBytesLen)
-
-		if minDocID != math.MaxUint64 && maxDocID != 0 {
-			stats.Store("vector_min_doc_id", fieldName, minDocID)
-			stats.Store("vector_max_doc_id", fieldName, maxDocID)
-		}
 	}
 }
diff --git a/faiss_vector_test.go b/faiss_vector_test.go
index b1b61fe..f1ad691 100644
--- a/faiss_vector_test.go
+++ b/faiss_vector_test.go
@@ -407,7 +407,7 @@ func TestVectorSegment(t *testing.T) {
 
 	fieldsSectionsMap := vecSegBase.fieldsSectionsMap
 	stubVecFieldStartAddr := fieldsSectionsMap[vecSegBase.fieldsMap["stubVec"]-1][SectionFaissVectorIndex]
-	docValueStart, docValueEnd, numVecs, _, _, _, indexBytesLen,
+	docValueStart, docValueEnd, numVecs, _, indexBytesLen,
 		_ := getVectorSectionContentOffsets(vecSegBase, stubVecFieldStartAddr)
 
 	if docValueStart != fieldNotUninverted {
diff --git a/section_faiss_vector_index.go b/section_faiss_vector_index.go
index 0bee8d4..95a90db 100644
--- a/section_faiss_vector_index.go
+++ b/section_faiss_vector_index.go
@@ -771,8 +771,6 @@ func getVectorSectionContentOffsets(sb *SegmentBase, offset uint64) (
 	docValueEnd,
 	numVecs,
 	vecDocIDsMappingOffset,
-	minDocID,
-	maxDocID,
 	indexBytesLen,
 	indexBytesOffset uint64,
 ) {
@@ -791,19 +789,15 @@ func getVectorSectionContentOffsets(sb *SegmentBase, offset uint64) (
 	pos += uint64(n)
 
 	vecDocIDsMappingOffset = pos
-	minDocID = uint64(math.MaxUint64)
-	maxDocID = uint64(0)
-	var docID uint64
+	// Just based on the number of vectors, we can't skip the vecID:docID mapping
+	// This is because the vectorIDs and docIDs are encoded in a variable length
+	// format, exact size of which varies based on the value of the ID.
+	// So, we need to iterate over the vectorID:docID mapping to determine the
+	// exact offset of the index bytes.
 	for i := 0; i < int(numVecs); i++ {
 		_, n := binary.Varint(sb.mem[pos : pos+binary.MaxVarintLen64])
 		pos += uint64(n)
-		docID, n = binary.Uvarint(sb.mem[pos : pos+binary.MaxVarintLen64])
-		if docID < minDocID {
-			minDocID = docID
-		}
-		if docID > maxDocID {
-			maxDocID = docID
-		}
+		_, n = binary.Uvarint(sb.mem[pos : pos+binary.MaxVarintLen64])
 		pos += uint64(n)
 	}