Enable early termination for HNSW by default

docs/changelog/130564.yaml

@@ -0,0 +1,5 @@
+pr: 130564
+summary: Enable early termination for HNSW by default
+area: Vector Search
+type: enhancement
+issues: []

docs/reference/elasticsearch/index-settings/index-modules.md

@@ -153,7 +153,7 @@ $$$index-refresh-interval-setting$$$
 :   How often to perform a refresh operation, which makes recent changes to the index visible to search. If this setting is not explicitly set, shards that haven’t seen search traffic for at least `index.search.idle.after` seconds will not receive background refreshes until they receive a search request. Searches that hit an idle shard where a refresh is pending will trigger a refresh as part of the search operation for that shard only. This behavior aims to automatically optimize bulk indexing in the default case when no searches are performed. To opt out of this behavior, set an explicit value for the refresh interval, even if it matches the default value.
 
     The value defaults to `1s` in {{stack}} and `5s` in {{serverless-short}}. In {{serverless-short}}, `5s` is also the minimum value that can be set.
-    
+
     In both cases, the setting can be set to `-1` to disable refresh.
 
 $$$index-max-result-window$$$
@@ -270,4 +270,4 @@ $$$index-esql-stored-fields-sequential-proportion$$$
 :   Tuning parameter for deciding when {{esql}} will load [stored fields](/reference/elasticsearch/rest-apis/retrieve-selected-fields.md#stored-fields) using a strategy tuned for loading dense sequence of documents. Allows values between 0.0 and 1.0 and defaults to 0.2. Indices with documents smaller than 10kb may see speed improvements loading `text` fields by setting this lower.
 
 $$$index-dense-vector-hnsw-early-termination$$$ `index.dense_vector.hnsw_early_termination` {applies_to}`stack: ga 9.2` {applies_to}`serverless: all`
-:   Whether to apply _patience_ based early termination strategy to knn queries over HNSW graphs (see [paper](https://cs.uwaterloo.ca/~jimmylin/publications/Teofili_Lin_ECIR2025.pdf)). This is only applicable to `dense_vector` fields with `hnsw`, `int8_hnsw`, `int4_hnsw` and `bbq_hnsw` index types. Defaults to `false`.
+:   Whether to apply _patience_ based early termination strategy to knn queries over HNSW graphs (see [paper](https://cs.uwaterloo.ca/~jimmylin/publications/Teofili_Lin_ECIR2025.pdf)). This is only applicable to `dense_vector` fields with `hnsw`, `int8_hnsw`, `int4_hnsw` and `bbq_hnsw` index types. Defaults to `true`.

server/src/internalClusterTest/java/org/elasticsearch/search/query/VectorIT.java

@@ -130,6 +130,11 @@ public void testFilteredQueryStrategy() {
     }
 
     public void testHnswEarlyTerminationQuery() {
+        client().admin()
+            .indices()
+            .prepareUpdateSettings(INDEX_NAME)
+            .setSettings(Settings.builder().put(DenseVectorFieldMapper.HNSW_EARLY_TERMINATION.getKey(), false))
+            .get();
         float[] vector = new float[16];
         randomVector(vector, 25);
         int upperLimit = 35;

server/src/main/java/org/elasticsearch/index/mapper/vectors/DenseVectorFieldMapper.java

@@ -125,7 +125,7 @@ public class DenseVectorFieldMapper extends FieldMapper {
     private static final float EPS = 1e-3f;
     public static final int BBQ_MIN_DIMS = 64;
 
-    private static final boolean DEFAULT_HNSW_EARLY_TERMINATION = false;
+    private static final boolean DEFAULT_HNSW_EARLY_TERMINATION = true;
 
     public static boolean isNotUnitVector(float magnitude) {
         return Math.abs(magnitude - 1.0f) > EPS;