feat: activate dormant paradigms — progressive looping, evolution with loss feedback, no progressive_unfreeze\n\nWith STE+AdamW (not MeZO), we can afford multi-loop training.\nProgressive loop schedule: 1→2→3 loops as training advances.\nEvolution engine now receives previous step loss for surprise\ndetection and memory writes.\nProgressive unfreeze disabled by default (counterproductive with backprop)."

chimera/training/hyper.py

@@ -119,10 +119,38 @@ class ProgressiveUnfreezer:
         return self._current
 
 
+class ProgressiveLoopScheduler:
+    """Gradually increase Parcae loop depth during training.
+
+    With STE+AdamW (not MeZO), multi-loop training is affordable.
+    Progressive schedule avoids instability from deep loops early on.
+
+    Default: loops=1 for 20%, loops=2 for 40%, loops=3 for 40%.
+    """
+
+    def __init__(self, total_steps: int, max_loops: int = 3):
+        self._total = total_steps
+        self._max_loops = max_loops
+        # Schedule: (fraction_done_threshold, num_loops)
+        self._schedule = [
+            (0.20, 1),  # First 20%: stabilize weights
+            (0.60, 2),  # Next 40%: learn to iterate
+            (1.01, min(3, max_loops)),  # Last 40%: deep refinement
+        ]
+
+    def get_loops(self, step: int) -> int:
+        frac = step / max(1, self._total)
+        for threshold, loops in self._schedule:
+            if frac < threshold:
+                return loops
+        return self._schedule[-1][1]
+
+
 def patch_training_loops(model, num_loops=1) -> None:
+    """Set initial loop config. Use ProgressiveLoopScheduler to change during training."""
     if hasattr(model, "loop_controller"):
         model.loop_controller.loop_default = num_loops
         model.loop_controller.loop_min = 1
-        model.loop_controller.loop_max = max(num_loops, 1)
+        model.loop_controller.loop_max = max(num_loops, 3)  # ← allow up to 3
     if hasattr(model, "evo_every_n_layers"):
         model.evo_every_n_layers = max(model.evo_every_n_layers, 8)