diff --git a/docs/source/contributor-guide/spark_expressions_support.md b/docs/source/contributor-guide/spark_expressions_support.md
index 65f941210a..d9897fe94c 100644
--- a/docs/source/contributor-guide/spark_expressions_support.md
+++ b/docs/source/contributor-guide/spark_expressions_support.md
@@ -40,6 +40,9 @@
 - [ ] approx_top_k_combine
 - [ ] array_agg
 - [x] avg
+  - Spark 3.4.3 (2026-05-26)
+  - Spark 3.5.8 (2026-05-26): aggregate logic identical to 3.4.3
+  - Spark 4.0.1 (2026-05-26): aggregate logic identical to 3.5.8; only `QueryContext` import path differs. `YearMonthIntervalType` and `DayTimeIntervalType` inputs (supported by Spark) fall back to Spark in Comet.
 - [x] bit_and
 - [x] bit_or
 - [x] bit_xor
diff --git a/spark/src/main/scala/org/apache/comet/serde/aggregates.scala b/spark/src/main/scala/org/apache/comet/serde/aggregates.scala
index 2714a7e466..a9ae740900 100644
--- a/spark/src/main/scala/org/apache/comet/serde/aggregates.scala
+++ b/spark/src/main/scala/org/apache/comet/serde/aggregates.scala
@@ -155,8 +155,8 @@ object CometCount extends CometAggregateExpressionSerde[Count] {
 
 object CometAverage extends CometAggregateExpressionSerde[Average] {
 
-  override def getIncompatibleReasons(): Seq[String] = Seq(
-    "Falls back to Spark in ANSI mode. Supports all numeric inputs except decimal types.")
+  override def getUnsupportedReasons(): Seq[String] = Seq(
+    "YearMonthIntervalType and DayTimeIntervalType inputs are not supported")
 
   override def convert(
       aggExpr: AggregateExpression,
diff --git a/spark/src/test/resources/sql-tests/expressions/aggregate/avg.sql b/spark/src/test/resources/sql-tests/expressions/aggregate/avg.sql
index c718a80ded..9d41d86367 100644
--- a/spark/src/test/resources/sql-tests/expressions/aggregate/avg.sql
+++ b/spark/src/test/resources/sql-tests/expressions/aggregate/avg.sql
@@ -26,3 +26,107 @@ SELECT avg(i), avg(l), avg(f), avg(d) FROM test_avg
 
 query tolerance=1e-6
 SELECT grp, avg(d) FROM test_avg GROUP BY grp ORDER BY grp
+
+-- single-row group (count == 1)
+query tolerance=1e-6
+SELECT grp, avg(d) FROM test_avg WHERE i = 3 GROUP BY grp
+
+-- byte and short input types
+statement
+CREATE TABLE test_avg_small(b tinyint, s smallint, grp string) USING parquet
+
+statement
+INSERT INTO test_avg_small VALUES (1, 100, 'a'), (2, 200, 'a'), (3, 300, 'b'), (NULL, NULL, 'b'), (-1, -100, 'a')
+
+query tolerance=1e-6
+SELECT avg(b), avg(s) FROM test_avg_small
+
+query tolerance=1e-6
+SELECT grp, avg(b), avg(s) FROM test_avg_small GROUP BY grp ORDER BY grp
+
+-- all-NULL input returns NULL
+statement
+CREATE TABLE test_avg_all_null(v double, grp string) USING parquet
+
+statement
+INSERT INTO test_avg_all_null VALUES (NULL, 'a'), (NULL, 'a'), (NULL, 'b')
+
+query
+SELECT avg(v) FROM test_avg_all_null
+
+query
+SELECT grp, avg(v) FROM test_avg_all_null GROUP BY grp ORDER BY grp
+
+-- empty input (no rows) returns NULL
+statement
+CREATE TABLE test_avg_empty(v double) USING parquet
+
+query
+SELECT avg(v) FROM test_avg_empty
+
+-- NaN and infinity input on doubles
+statement
+CREATE TABLE test_avg_special(v double, grp string) USING parquet
+
+statement
+INSERT INTO test_avg_special VALUES
+  (double('NaN'), 'nan_only'),
+  (1.0, 'nan_only'),
+  (double('Infinity'), 'pos_inf_only'),
+  (1.0, 'pos_inf_only'),
+  (double('Infinity'), 'mixed_inf'),
+  (double('-Infinity'), 'mixed_inf'),
+  (double('-Infinity'), 'neg_inf_only'),
+  (-2.0, 'neg_inf_only')
+
+query tolerance=1e-6
+SELECT grp, avg(v) FROM test_avg_special GROUP BY grp ORDER BY grp
+
+-- boundary integer values
+statement
+CREATE TABLE test_avg_bounds(l long, grp string) USING parquet
+
+statement
+INSERT INTO test_avg_bounds VALUES
+  (9223372036854775807, 'maxes'),
+  (9223372036854775807, 'maxes'),
+  (-9223372036854775808, 'mins'),
+  (-9223372036854775808, 'mins'),
+  (9223372036854775807, 'mixed'),
+  (-9223372036854775808, 'mixed')
+
+query tolerance=1e-6
+SELECT grp, avg(l) FROM test_avg_bounds GROUP BY grp ORDER BY grp
+
+-- negative-only inputs
+statement
+CREATE TABLE test_avg_negative(d double) USING parquet
+
+statement
+INSERT INTO test_avg_negative VALUES (-1.5), (-2.5), (-3.5), (-0.0)
+
+query tolerance=1e-6
+SELECT avg(d) FROM test_avg_negative
+
+-- decimal column at higher precision
+statement
+CREATE TABLE test_avg_decimal(d decimal(20, 5), grp string) USING parquet
+
+statement
+INSERT INTO test_avg_decimal VALUES
+  (10.50000, 'a'),
+  (20.25000, 'a'),
+  (NULL, 'a'),
+  (-5.00000, 'b'),
+  (0.00000, 'b'),
+  (5.00000, 'b')
+
+query
+SELECT avg(d) FROM test_avg_decimal
+
+query
+SELECT grp, avg(d) FROM test_avg_decimal GROUP BY grp ORDER BY grp
+
+-- count(*) and avg in the same query for cross-check
+query tolerance=1e-6
+SELECT count(d), avg(d) FROM test_avg_decimal