diff --git a/tests/utils/reward_score/test_math_dapo_on_cpu.py b/tests/utils/reward_score/test_math_dapo_on_cpu.py
new file mode 100644
index 00000000000..19c6ef6191f
--- /dev/null
+++ b/tests/utils/reward_score/test_math_dapo_on_cpu.py
@@ -0,0 +1,71 @@
+# Copyright 2026 Bytedance Ltd. and/or its affiliates
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from verl.utils.reward_score.math_dapo import compute_score
+
+
+def _boxed(answer: str) -> str:
+    return f"\\boxed{{{answer}}}"
+
+
+def _documented_length_answer() -> tuple[str, str]:
+    answer = "x" * 151
+    solution = _boxed(answer)
+    assert len(solution) == 159
+    return solution, answer
+
+
+def _assert_correct(result: dict, answer: str) -> None:
+    assert result == {"score": 1.0, "acc": True, "pred": answer}
+
+
+def test_documented_length_boxed_answer_in_strict_mode():
+    solution, answer = _documented_length_answer()
+
+    result = compute_score(solution, answer, strict_box_verify=True)
+
+    _assert_correct(result, answer)
+
+
+def test_documented_length_boxed_answer_in_fallback_mode():
+    solution, answer = _documented_length_answer()
+
+    result = compute_score(solution, answer)
+
+    _assert_correct(result, answer)
+
+
+def test_strict_box_uses_the_final_box():
+    result = compute_score(r"\boxed{99} more reasoning \boxed{42}", "42", strict_box_verify=True)
+
+    _assert_correct(result, "42")
+
+
+def test_pause_token_indices_do_not_shorten_the_character_window():
+    solution, answer = _documented_length_answer()
+
+    result = compute_score(
+        solution,
+        answer,
+        strict_box_verify=True,
+        pause_tokens_index=[0, 1, 2, len(solution)],
+    )
+
+    _assert_correct(result, answer)
+
+
+def test_strict_box_without_a_box_remains_incorrect():
+    result = compute_score("reasoning only", "42", strict_box_verify=True)
+
+    assert result == {"score": -1.0, "acc": False, "pred": None}
diff --git a/verl/utils/reward_score/math_dapo.py b/verl/utils/reward_score/math_dapo.py
index ba652f99cc1..05b1482c085 100644
--- a/verl/utils/reward_score/math_dapo.py
+++ b/verl/utils/reward_score/math_dapo.py
@@ -206,9 +206,6 @@ def is_correct_strict_box(
     # Extract the relevant part of the prediction
     if pause_tokens_index is not None:
         assert len(pause_tokens_index) == 4
-        pred = pred[pause_tokens_index[-1] - 100 :]
-    else:
-        pred = pred[-100:]
 
     # Extract and check the boxed answer
     boxed_pred = last_boxed_only_string(pred)