diff --git a/btf/strings.go b/btf/strings.go
index f8ae08cfb..114d25001 100644
--- a/btf/strings.go
+++ b/btf/strings.go
@@ -15,6 +15,7 @@ import (
 type stringTable struct {
 	base    *stringTable
 	offsets []uint32
+	prevIdx int
 	strings []string
 }
 
@@ -61,7 +62,7 @@ func readStringTable(r sizedReader, base *stringTable) (*stringTable, error) {
 		return nil, errors.New("first item in string table is non-empty")
 	}
 
-	return &stringTable{base, offsets, strings}, nil
+	return &stringTable{base, offsets, 0, strings}, nil
 }
 
 func splitNull(data []byte, atEOF bool) (advance int, token []byte, err error) {
@@ -84,15 +85,28 @@ func (st *stringTable) Lookup(offset uint32) (string, error) {
 }
 
 func (st *stringTable) lookup(offset uint32) (string, error) {
+	// Fast path: zero offset is the empty string, looked up frequently.
 	if offset == 0 && st.base == nil {
 		return "", nil
 	}
 
+	// Accesses tend to be globally increasing, so check if the next string is
+	// the one we want. This skips the binary search in about 50% of cases.
+	if st.prevIdx+1 < len(st.offsets) && st.offsets[st.prevIdx+1] == offset {
+		st.prevIdx++
+		return st.strings[st.prevIdx], nil
+	}
+
 	i, found := slices.BinarySearch(st.offsets, offset)
 	if !found {
 		return "", fmt.Errorf("offset %d isn't start of a string", offset)
 	}
 
+	// Set the new increment index, but only if its greater than the current.
+	if i > st.prevIdx+1 {
+		st.prevIdx = i
+	}
+
 	return st.strings[i], nil
 }
 
diff --git a/btf/strings_test.go b/btf/strings_test.go
index d34726d49..e97c62672 100644
--- a/btf/strings_test.go
+++ b/btf/strings_test.go
@@ -112,5 +112,5 @@ func newStringTable(strings ...string) *stringTable {
 		offset += uint32(len(str)) + 1 // account for NUL
 	}
 
-	return &stringTable{nil, offsets, strings}
+	return &stringTable{nil, offsets, 0, strings}
 }