diff --git a/motrix_envs/src/motrix_envs/locomotion/go1/dreamwaq.py b/motrix_envs/src/motrix_envs/locomotion/go1/dreamwaq.py
index 8a1295f..c3086fe 100644
--- a/motrix_envs/src/motrix_envs/locomotion/go1/dreamwaq.py
+++ b/motrix_envs/src/motrix_envs/locomotion/go1/dreamwaq.py
@@ -118,8 +118,8 @@ class DreamWaQCfg(Go1WalkNpEnvCfg):
         r = self.reward_config.scales
         r.clear()
         r.update({
-            "tracking_lin_vel": 1.0,
-            "tracking_ang_vel": 0.5,
+            "tracking_lin_vel": 1.5,
+            "tracking_ang_vel": 1.0,
             "lin_vel_z": -2.0,
             "ang_vel_xy": -0.05,
             "orientation": -0.2,
@@ -429,7 +429,7 @@ class DreamWaQTask(Go1WalkTask):
                     cy = half_y - self._border - row * self._cell_size - self._cell_size / 2
                     all_origins[row, col] = [cx, cy]
             self._terrain_origins = all_origins
-            self._max_init_level = 3  # 先 0-3，适应后再提到 5
+            self._max_init_level = 5  # 先 0-3，适应后再提到 5
 
         if num_reset > 0 and self._init_done and state is not None and hasattr(state, 'info'):
             old_info = state.info
diff --git a/motrix_rl/src/motrix_rl/tasks/go1.py b/motrix_rl/src/motrix_rl/tasks/go1.py
index ae90a47..fea4da5 100644
--- a/motrix_rl/src/motrix_rl/tasks/go1.py
+++ b/motrix_rl/src/motrix_rl/tasks/go1.py
@@ -112,7 +112,7 @@ class rslrl:
 
             # Runner 设置（严格对齐上游 LeggedRobotCfgPPO + Go1RoughCfgPPO）
             runner.seed = 5                     # 上游 seed=5
-            runner.max_iterations = 1000  # 论文原值
+            runner.max_iterations = 3000  # 续训到 3000 轮
             runner.num_steps_per_env = 24
             runner.experiment_name = "go1_dreamwaq_walk"
             runner.save_interval = 50