diff --git a/ml/src/labeling/meta_labeling/secondary_model.rs b/ml/src/labeling/meta_labeling/secondary_model.rs
index e28598d6f..6332d5a7c 100644
--- a/ml/src/labeling/meta_labeling/secondary_model.rs
+++ b/ml/src/labeling/meta_labeling/secondary_model.rs
@@ -145,6 +145,10 @@ pub struct SecondaryBettingModel {
     total_predictions: Arc<AtomicU64>,
     total_trades: Arc<AtomicU64>,
     total_bet_size: Arc<AtomicU64>, // Stored as fixed-point (multiply by 1e6)
+    /// Exponential moving average of prediction confidence, fixed-point (x 1_000_000)
+    confidence_ema: Arc<AtomicU64>,
+    /// EMA smoothing factor (higher = more weight on recent observations)
+    confidence_ema_alpha: f64,
 }
 
 impl SecondaryBettingModel {
@@ -157,6 +161,8 @@ impl SecondaryBettingModel {
             total_predictions: Arc::new(AtomicU64::new(0)),
             total_trades: Arc::new(AtomicU64::new(0)),
             total_bet_size: Arc::new(AtomicU64::new(0)),
+            confidence_ema: Arc::new(AtomicU64::new(0)),
+            confidence_ema_alpha: 0.99,
         })
     }
 
@@ -188,6 +194,21 @@ impl SecondaryBettingModel {
         // Update statistics
         self.total_predictions.fetch_add(1, Ordering::Relaxed);
 
+        // Update confidence EMA (fixed-point encoding: value * 1_000_000)
+        {
+            let current_ema_fixed = self.confidence_ema.load(Ordering::Relaxed);
+            let current_ema = current_ema_fixed as f64 / 1_000_000.0;
+            let alpha = self.confidence_ema_alpha;
+            let new_ema = if current_ema_fixed == 0 && self.total_predictions.load(Ordering::Relaxed) == 1 {
+                // First prediction: initialize EMA to this confidence value
+                primary.confidence
+            } else {
+                alpha * current_ema + (1.0 - alpha) * primary.confidence
+            };
+            self.confidence_ema
+                .store((new_ema * 1_000_000.0) as u64, Ordering::Relaxed);
+        }
+
         // Validate inputs
         if primary.features.is_empty() || features.is_empty() {
             return Err(MLError::ValidationError {
@@ -335,7 +356,7 @@ impl SecondaryBettingModel {
             total_trades,
             total_rejections: total_predictions.saturating_sub(total_trades),
             average_bet_size,
-            average_confidence: 0.0, // TODO: Track confidence running average
+            average_confidence: self.confidence_ema.load(Ordering::Relaxed) as f64 / 1_000_000.0,
         }
     }
 
@@ -344,6 +365,7 @@ impl SecondaryBettingModel {
         self.total_predictions.store(0, Ordering::Relaxed);
         self.total_trades.store(0, Ordering::Relaxed);
         self.total_bet_size.store(0, Ordering::Relaxed);
+        self.confidence_ema.store(0, Ordering::Relaxed);
     }
 }
 
@@ -394,6 +416,42 @@ mod tests {
         assert!(combined < 0.6);
     }
 
+    #[test]
+    fn test_secondary_model_confidence_ema() {
+        let config = SecondaryModelConfig::default();
+        let mut model = SecondaryBettingModel::new(config).unwrap();
+
+        let primary = PrimaryPrediction {
+            direction: 1,
+            confidence: 0.8,
+            expected_return: 0.05,
+            features: vec![0.5, 0.6, 0.7],
+        };
+        let features = vec![0.2, 0.8, 0.7];
+
+        // First prediction initializes EMA to 0.8
+        let _ = model.should_trade(&primary, &features).unwrap();
+        let stats = model.get_statistics();
+        assert!((stats.average_confidence - 0.8).abs() < 0.01);
+
+        // Second prediction with confidence 0.6
+        // EMA = 0.99 * 0.8 + 0.01 * 0.6 = 0.798
+        let primary2 = PrimaryPrediction {
+            direction: 1,
+            confidence: 0.6,
+            expected_return: 0.05,
+            features: vec![0.5, 0.6, 0.7],
+        };
+        let _ = model.should_trade(&primary2, &features).unwrap();
+        let stats2 = model.get_statistics();
+        assert!((stats2.average_confidence - 0.798).abs() < 0.01);
+
+        // Reset clears EMA
+        model.reset_statistics();
+        let stats3 = model.get_statistics();
+        assert!((stats3.average_confidence - 0.0).abs() < 1e-6);
+    }
+
     #[test]
     fn test_bet_size_calculation() {
         let config = SecondaryModelConfig::default();
diff --git a/ml/src/ppo/action_masking.rs b/ml/src/ppo/action_masking.rs
index 6ea2ee296..04fd58963 100644
--- a/ml/src/ppo/action_masking.rs
+++ b/ml/src/ppo/action_masking.rs
@@ -76,14 +76,32 @@ pub fn create_action_mask(current_position: f64, max_position: f64, num_actions:
         }
 
         // HOLD (action 2) is always valid
-    } else {
-        // Future 45-action implementation (Phase 3)
-        // This will be implemented when expanding to factored action space
-        // For now, default all to valid
-        // TODO(Phase 3): Implement factored action masking
-        // - Map action index → FactoredAction
-        // - Get target_exposure from action
-        // - Mask if |target_exposure| > max_position
+    } else if num_actions == 45 {
+        // Factored 45-action space: (direction: 3) x (size: 5) x (urgency: 3) = 45
+        // direction = action_idx / 15: 0=Buy, 1=Sell, 2=Hold
+        for action_idx in 0..num_actions {
+            let direction = action_idx / 15;
+            match direction {
+                0 => {
+                    // Buy: mask if at or above max position
+                    if current_position >= max_position {
+                        if let Some(m) = mask.get_mut(action_idx) {
+                            *m = false;
+                        }
+                    }
+                }
+                1 => {
+                    // Sell: mask if at or below negative max position
+                    if current_position <= -max_position {
+                        if let Some(m) = mask.get_mut(action_idx) {
+                            *m = false;
+                        }
+                    }
+                }
+                // 2 = Hold: always valid, no masking needed
+                _ => {}
+            }
+        }
     }
 
     mask
@@ -192,6 +210,70 @@ mod tests {
         assert_eq!(values[2], 0.2);
     }
 
+    #[test]
+    fn test_create_action_mask_45_flat_position() {
+        // Flat position: all 45 actions valid
+        let mask = create_action_mask(0.0, 2.0, 45);
+        assert_eq!(mask.len(), 45);
+        assert!(mask.iter().all(|&v| v));
+    }
+
+    #[test]
+    fn test_create_action_mask_45_at_max_position() {
+        // At max long: Buy actions (0..15) masked, Sell (15..30) and Hold (30..45) valid
+        let mask = create_action_mask(2.0, 2.0, 45);
+        assert_eq!(mask.len(), 45);
+        // Buy actions (direction=0, indices 0..15) should be masked
+        for i in 0..15 {
+            assert_eq!(
+                mask.get(i).copied().unwrap_or(true),
+                false,
+                "Buy action {} should be masked at max position",
+                i
+            );
+        }
+        // Sell actions (direction=1, indices 15..30) should be valid
+        for i in 15..30 {
+            assert_eq!(
+                mask.get(i).copied().unwrap_or(false),
+                true,
+                "Sell action {} should be valid at max position",
+                i
+            );
+        }
+        // Hold actions (direction=2, indices 30..45) should be valid
+        for i in 30..45 {
+            assert_eq!(
+                mask.get(i).copied().unwrap_or(false),
+                true,
+                "Hold action {} should be valid at max position",
+                i
+            );
+        }
+    }
+
+    #[test]
+    fn test_create_action_mask_45_at_min_position() {
+        // At max short: Sell actions (15..30) masked, Buy (0..15) and Hold (30..45) valid
+        let mask = create_action_mask(-2.0, 2.0, 45);
+        for i in 0..15 {
+            assert_eq!(mask.get(i).copied().unwrap_or(false), true);
+        }
+        for i in 15..30 {
+            assert_eq!(mask.get(i).copied().unwrap_or(true), false);
+        }
+        for i in 30..45 {
+            assert_eq!(mask.get(i).copied().unwrap_or(false), true);
+        }
+    }
+
+    #[test]
+    fn test_create_action_mask_45_partial_position() {
+        // Partial position: neither at max nor min, all valid
+        let mask = create_action_mask(1.0, 2.0, 45);
+        assert!(mask.iter().all(|&v| v));
+    }
+
     #[test]
     fn test_apply_mask_all_invalid() {
         let device = Device::Cpu;
diff --git a/ml/src/ppo/continuous_action_masking.rs b/ml/src/ppo/continuous_action_masking.rs
index 02b0ae4e2..e3851721d 100644
--- a/ml/src/ppo/continuous_action_masking.rs
+++ b/ml/src/ppo/continuous_action_masking.rs
@@ -22,7 +22,7 @@
 //! - Penalty = penalty_coeff × max(0, |action| - soft_threshold)²
 //! - Encourages staying away from hard limits (margin of safety)
 
-use candle_core::{Device, Tensor};
+use candle_core::{Device, IndexOp, Tensor};
 use serde::{Deserialize, Serialize};
 
 use crate::MLError;
@@ -90,12 +90,22 @@ impl ContinuousActionConstraints {
             )));
         }
 
-        // TODO(Phase 2): Extract current position from state to compute dynamic bounds
-        // Example: let current_position = state.get(0)?.get(0)?.to_scalar::<f32>()?;
+        // Extract current position from state tensor (first element of first sample).
+        // Safe access: returns 0.0 if tensor indexing fails.
+        let current_position = state
+            .i((0, 0))
+            .and_then(|t| t.to_scalar::<f32>())
+            .unwrap_or(0.0);
+
+        // Compute asymmetric bounds based on current position:
+        // - max_long: remaining room to go long
+        // - max_short: remaining room to go short
+        let max_long = max_position_abs - current_position.max(0.0);
+        let max_short = max_position_abs + current_position.min(0.0);
 
         Ok(Self {
-            min_position: -max_position_abs,
-            max_position: max_position_abs,
+            min_position: -max_short,
+            max_position: max_long,
             penalty_coeff: 1.0,               // Moderate penalty
             soft_threshold_fraction: 0.8,     // Penalize beyond 80% of limit
         })
@@ -570,6 +580,54 @@ mod tests {
         assert!(!constraints.is_within_soft_bounds(2.0));
     }
 
+    #[test]
+    fn test_from_state_asymmetric_bounds_long() {
+        let device = Device::Cpu;
+        // State with current_position = 1.5 in first element
+        let mut state_data = vec![0.0f32; 64];
+        state_data[0] = 1.5;
+        let state = Tensor::new(state_data.as_slice(), &device)
+            .unwrap()
+            .reshape((1, 64))
+            .unwrap();
+        let constraints = ContinuousActionConstraints::from_state(&state, 2.0).unwrap();
+
+        // max_long = 2.0 - max(1.5, 0.0) = 0.5
+        assert!((constraints.max_position - 0.5).abs() < 1e-6);
+        // max_short = 2.0 + min(1.5, 0.0) = 2.0, so min_position = -2.0
+        assert!((constraints.min_position - (-2.0)).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_from_state_asymmetric_bounds_short() {
+        let device = Device::Cpu;
+        // State with current_position = -1.0
+        let mut state_data = vec![0.0f32; 64];
+        state_data[0] = -1.0;
+        let state = Tensor::new(state_data.as_slice(), &device)
+            .unwrap()
+            .reshape((1, 64))
+            .unwrap();
+        let constraints = ContinuousActionConstraints::from_state(&state, 2.0).unwrap();
+
+        // max_long = 2.0 - max(-1.0, 0.0) = 2.0
+        assert!((constraints.max_position - 2.0).abs() < 1e-6);
+        // max_short = 2.0 + min(-1.0, 0.0) = 1.0, so min_position = -1.0
+        assert!((constraints.min_position - (-1.0)).abs() < 1e-6);
+    }
+
+    #[test]
+    fn test_from_state_flat_position_symmetric() {
+        let device = Device::Cpu;
+        // State with current_position = 0.0 (flat)
+        let state = Tensor::zeros((1, 64), DType::F32, &device).unwrap();
+        let constraints = ContinuousActionConstraints::from_state(&state, 2.0).unwrap();
+
+        // Symmetric: max_long = 2.0, max_short = 2.0
+        assert!((constraints.max_position - 2.0).abs() < 1e-6);
+        assert!((constraints.min_position - (-2.0)).abs() < 1e-6);
+    }
+
     #[test]
     fn test_mask_continuous_actions_integration() {
         let device = Device::Cpu;
diff --git a/ml/src/safety/memory_manager.rs b/ml/src/safety/memory_manager.rs
index 60dc3b457..9dbca86b7 100644
--- a/ml/src/safety/memory_manager.rs
+++ b/ml/src/safety/memory_manager.rs
@@ -358,15 +358,9 @@ impl SafeMemoryManager {
             usage.reset_peak();
         }
 
-        // Force garbage collection hint (if applicable)
-        // Note: Rust does not have a standard garbage collector
-        // This is a placeholder for future integration with alternative GC implementations
-        #[cfg(feature = "gc")]
-        {
-            // TODO: Integrate with a Rust GC library like `gc` or `rust-gc` if needed
-            // For now, this is a no-op as Rust uses RAII and ownership for memory management
-            tracing::debug!("GC hint requested but no GC is available in standard Rust");
-        }
+        // No garbage collector needed: Rust's ownership model and RAII handle
+        // deallocation automatically when Tensors and VarMaps go out of scope.
+        // GPU memory (CUDA/Metal) is freed when candle Tensors are dropped.
 
         info!("Memory cleanup completed for device: {}", device_key);
         Ok(())