diff --git a/configs.yaml b/configs.yaml
index f5c99b2..1a06411 100644
--- a/configs.yaml
+++ b/configs.yaml
@@ -47,7 +47,7 @@ defaults:
   decoder:
     {mlp_keys: '$^', cnn_keys: 'image', act: 'SiLU', norm: True, cnn_depth: 32, kernel_size: 4, minres: 4, mlp_layers: 2, mlp_units: 512, cnn_sigmoid: False, image_dist: mse, vector_dist: symlog_mse, outscale: 1.0}
   actor:
-    {layers: 2, dist: 'normal', entropy: 3e-4, unimix_ratio: 0.01, min_std: 0.1, max_std: 1.0, temp: 0.1, lr: 3e-5, eps: 1e-5, grad_clip: 100.0, outscale: 1.0}
+    {layers: 2, dist: 'normal', entropy: 3e-4, unimix_ratio: 0.01, std: 'learned', min_std: 0.1, max_std: 1.0, temp: 0.1, lr: 3e-5, eps: 1e-5, grad_clip: 100.0, outscale: 1.0}
   critic:
     {layers: 2, dist: 'symlog_disc', slow_target: True, slow_target_update: 1, slow_target_fraction: 0.02, lr: 3e-5, eps: 1e-5, grad_clip: 100.0, outscale: 0.0}
   reward_head:
@@ -123,7 +123,7 @@ crafter:
   units: 1024
   encoder: {mlp_keys: '$^', cnn_keys: 'image', cnn_depth: 96, mlp_layers: 5, mlp_units: 1024}
   decoder: {mlp_keys: '$^', cnn_keys: 'image', cnn_depth: 96, mlp_layers: 5, mlp_units: 1024}
-  actor: {layers: 5, dist: 'onehot'}
+  actor: {layers: 5, dist: 'onehot', std: 'none'}
   value: {layers: 5}
   reward_head: {layers: 5}
   cont_head: {layers: 5}
@@ -136,7 +136,7 @@ atari100k:
   train_ratio: 1024
   video_pred_log: true
   eval_episode_num: 100
-  actor: {dist: 'onehot'}
+  actor: {dist: 'onehot', std: 'none'}
   imag_gradient: 'reinforce'
   stickey: False
   lives: unused
@@ -161,7 +161,7 @@ minecraft:
   units: 1024
   encoder: {mlp_keys: 'inventory|inventory_max|equipped|health|hunger|breath|obs_reward', cnn_keys: 'image', cnn_depth: 96, mlp_layers: 5, mlp_units: 1024}
   decoder: {mlp_keys: 'inventory|inventory_max|equipped|health|hunger|breath', cnn_keys: 'image', cnn_depth: 96, mlp_layers: 5, mlp_units: 1024}
-  actor: {layers: 5, dist: 'onehot'}
+  actor: {layers: 5, dist: 'onehot', std: 'none'}
   value: {layers: 5}
   reward_head: {layers: 5}
   cont_head: {layers: 5}
@@ -172,7 +172,7 @@ minecraft:
 memorymaze:
   steps: 1e8
   action_repeat: 2
-  actor: {dist: 'onehot'}
+  actor: {dist: 'onehot', std: 'none'}
   imag_gradient: 'reinforce'
   task: 'memorymaze_9x9'
 
diff --git a/models.py b/models.py
index 6aed023..f0d3f63 100644
--- a/models.py
+++ b/models.py
@@ -228,7 +228,7 @@ class ImagBehavior(nn.Module):
             config.act,
             config.norm,
             config.actor["dist"],
-            "learned",
+            config.actor["std"],
             config.actor["min_std"],
             config.actor["max_std"],
             absmax=1.0,