diff --git a/crates/ml/src/cuda_pipeline/experience_kernels.cu b/crates/ml/src/cuda_pipeline/experience_kernels.cu
index 875ce46d3..92228e750 100644
--- a/crates/ml/src/cuda_pipeline/experience_kernels.cu
+++ b/crates/ml/src/cuda_pipeline/experience_kernels.cu
@@ -804,10 +804,12 @@ extern "C" __global__ void experience_action_select(
  *   3. Computes target position = exposure_fraction * max_position.
  *   4. Applies position adjustment delta with volatility-scaled tx cost.
  *   5. Runs dynamic trailing stop (regime-adaptive).
- *   6. Computes reward v6:
+ *   6. Computes reward v7:
  *        sparse = trade_return * patience_mult  (at trade exit, primary signal)
  *        reward = sparse  (Sparse Trade-Completion Only)
- *        if reward < 0: reward *= loss_aversion  (prospect theory)
+ *        + CEA bonus (counterfactual exposure advantage)
+ *        + order_credit (microstructure credit for order type selection)
+ *        + risk_efficiency (intra-trade risk-adjusted return bonus)
  *   7. Writes (batch_states, action, reward, done) to output replay buffer.
  *   8. Updates portfolio_states[0..19] in place.
  *   9. Increments current_timesteps[i].
@@ -836,7 +838,7 @@ extern "C" __global__ void experience_env_step(
     const float* __restrict__ batch_states,  /* #30: f32 states from state_gather */
     float max_position,
     float tx_cost_multiplier,
-    float loss_aversion,
+    float cea_weight,               /* v7: counterfactual exposure advantage weight */
     const __nv_bfloat16* __restrict__ features,
     int market_dim,
     int L,
@@ -856,15 +858,14 @@ extern "C" __global__ void experience_env_step(
     float margin_pct,           /* e.g. 0.06 (6% initial margin) */
     float dd_threshold,         /* drawdown fraction before penalty (0.02 = 2%) */
     float w_dd,                 /* drawdown penalty weight (1.0 = full) */
-    float beta_penalty,         /* anti-correlation: penalize market-aligned returns (0.0 = off) */
-    float trade_clustering_penalty, /* #25 trade clustering: CV(inter-trade times) * weight (0=off) */
+    float order_credit_weight,      /* v7: order type microstructure credit weight */
+    float risk_efficiency_weight,   /* v7: intra-trade risk efficiency weight */
     int mirror_active,              /* #10 mirror universe: invert exposure actions */
     /* #19 Position entropy: per-episode position visit histogram [N, 9].
      * NULL = disabled. Incremented at each timestep. Entropy bonus computed
      * at episode end (done=1) and added to final reward. */
     float* __restrict__ position_histogram,
     float position_entropy_weight,  /* #19: reward += weight * H(histogram). 0=disabled. */
-    float regret_blend,             /* #17: blend factor for counterfactual regret (0=pure PnL, 1=pure regret) */
     /* #33 Per-episode saboteur params [N, 3]: (spread_mult, fill_prob, slippage_mult).
      * NULL = disabled (use global scalars). When non-NULL, overrides
      * spread_cost, fill_ioc_fill_prob, tx_cost_multiplier per episode. */
diff --git a/crates/ml/src/cuda_pipeline/gpu_experience_collector.rs b/crates/ml/src/cuda_pipeline/gpu_experience_collector.rs
index 8f319bff6..e79b44de4 100644
--- a/crates/ml/src/cuda_pipeline/gpu_experience_collector.rs
+++ b/crates/ml/src/cuda_pipeline/gpu_experience_collector.rs
@@ -1787,7 +1787,7 @@ impl GpuExperienceCollector {
             // ── 5. Environment step (reward v5: trade-aware hybrid) ──────
             // max_pos and min_hold_bars_i32 already defined above (action_select block)
             let tx_cost = config.tx_cost_multiplier;
-            let rw_loss_av = config.loss_aversion;
+            let cea_w = config.cea_weight as f32;
             let l_i32 = timesteps as i32;
             // Task 8: current_t is now GPU-resident — pass device pointer
             // instead of host scalar.  step_counter_gpu[0] == t at this point.
@@ -1806,7 +1806,7 @@ impl GpuExperienceCollector {
                     .arg(&self.batch_states)
                     .arg(&max_pos)
                     .arg(&tx_cost)
-                    .arg(&rw_loss_av)
+                    .arg(&cea_w)
                     .arg(market_features_buf)
                     .arg(&md)
                     .arg(&l_i32)
@@ -1826,12 +1826,11 @@ impl GpuExperienceCollector {
                     .arg(&config.margin_pct)          // initial margin fraction (e.g. 0.06 = 6%)
                     .arg(&config.dd_threshold)        // drawdown threshold before penalty (0.02 = 2%)
                     .arg(&config.w_dd)                // drawdown penalty weight
-                    .arg(&config.beta_penalty)        // anti-correlation beta penalty
-                    .arg(&config.trade_clustering_penalty) // #25 trade clustering penalty
+                    .arg(&(config.order_credit_weight as f32))    // v7: order microstructure credit
+                    .arg(&(config.risk_efficiency_weight as f32))  // v7: risk efficiency bonus
                     .arg(&mirror_i32)                      // #10 mirror universe
                     .arg(&mut self.position_histogram)     // #19 position entropy histogram
                     .arg(&config.position_entropy_weight)  // #19 position entropy weight
-                    .arg(&config.regret_blend)             // #17 counterfactual regret
                     // #33 Per-episode saboteur params (0 = NULL = disabled)
                     .arg(&{
                         if self.saboteur_active {