diff --git a/docs/source/pi052.mdx b/docs/source/pi052.mdx
index 43fd8c918..73ccd6236 100644
--- a/docs/source/pi052.mdx
+++ b/docs/source/pi052.mdx
@@ -111,7 +111,7 @@ the expected prompt, text target, and action endpoints before scaling up.
 | -------------------------------- | -------------------------: | ------------------------------------------------------------------ |
 | `policy.recipe_path`             | `recipes/subtask_mem.yaml` | Selects the language/action objective mixture                      |
 | `policy.text_loss_weight`        |                      `1.0` | Language-head cross-entropy weight; `0` disables text training     |
-| `policy.flow_loss_weight`        |                      `5.0` | Continuous action flow-loss weight                                 |
+| `policy.flow_loss_weight`        |                     `10.0` | Continuous action flow-loss weight                                 |
 | `policy.enable_fast_action_loss` |                     `true` | Adds discrete FAST action-token supervision                        |
 | `policy.fast_action_loss_weight` |                      `1.0` | FAST cross-entropy weight                                          |
 | `policy.knowledge_insulation`    |                     `true` | Blocks action-loss gradients through the VLM K/V path              |
diff --git a/src/lerobot/policies/pi052/configuration_pi052.py b/src/lerobot/policies/pi052/configuration_pi052.py
index 6442c33d1..1d41a1f7c 100644
--- a/src/lerobot/policies/pi052/configuration_pi052.py
+++ b/src/lerobot/policies/pi052/configuration_pi052.py
@@ -48,11 +48,8 @@ class PI052Config(PI05Config):
     """Weight on the LM-head cross-entropy term. Set to ``0`` to disable
     text training entirely (reverts to flow-only / π0.5 behaviour)."""
 
-    flow_loss_weight: float = 5.0
-    """Weight on the action-expert flow-matching term. ``5.0`` — a milder
-    flow:text split than the paper's α=10, since the flow-only
-    ``low_level`` recipe already gives the action expert frequent
-    gradient. Lower it further if the LM head still underfits."""
+    flow_loss_weight: float = 10.0
+    """Weight on action-expert flow matching relative to text supervision."""
 
     # Backbone training ---------------------------------------------------
     unfreeze_lm_head: bool = True