diff --git a/docs/plans/2026-03-11-cuda-backtest-gpu-residency-plan.md b/docs/plans/2026-03-11-cuda-backtest-gpu-residency-plan.md
index 198751506..27b089e99 100644
--- a/docs/plans/2026-03-11-cuda-backtest-gpu-residency-plan.md
+++ b/docs/plans/2026-03-11-cuda-backtest-gpu-residency-plan.md
@@ -2028,16 +2028,16 @@ bitonic sort. Single download of extended WindowMetrics struct."
 Replace CPU backtest path in the standalone evaluation binary.
 
 **Files:**
-- Modify: `bin/fxt/src/commands/evaluate_baseline.rs` (the evaluate_baseline binary entry point)
-- Test: `SQLX_OFFLINE=true cargo check -p fxt --features cuda`
+- Modify: `crates/ml/examples/evaluate_baseline.rs` (declared as `[[example]]` in `crates/ml/Cargo.toml:230`)
+- Test: `SQLX_OFFLINE=true cargo check -p ml --example evaluate_baseline`
 
 - [ ] **Step 1: Add GPU evaluation path to evaluate_baseline**
 
-In `bin/fxt/src/commands/evaluate_baseline.rs`, at the evaluation entry point where the
+In `crates/ml/examples/evaluate_baseline.rs`, at the evaluation entry point where the
 `EvaluationEngine` is currently constructed, add a GPU path before the CPU path:
 
 ```rust
-use crate::cuda_pipeline::gpu_backtest_evaluator::{GpuBacktestEvaluator, GpuBacktestConfig};
+use ml::cuda_pipeline::gpu_backtest_evaluator::{GpuBacktestEvaluator, GpuBacktestConfig};
 
 // Try GPU evaluation first
 #[cfg(feature = "cuda")]
@@ -2076,13 +2076,13 @@ if let Ok(device) = Device::new_cuda(0) {
 
 - [ ] **Step 2: Run check**
 
-Run: `SQLX_OFFLINE=true cargo check -p fxt`
+Run: `SQLX_OFFLINE=true cargo check -p ml --example evaluate_baseline`
 Expected: PASS (compiles with and without cuda feature)
 
 - [ ] **Step 3: Commit**
 
 ```bash
-git add bin/fxt/src/commands/evaluate_baseline.rs
+git add crates/ml/examples/evaluate_baseline.rs
 git commit -m "feat(eval): wire GPU backtester into evaluate_baseline binary
 
 Standalone evaluation now uses CUDA backtest kernel when available.
@@ -2109,6 +2109,8 @@ use ml::cuda_pipeline::gpu_backtest_evaluator::{
     GpuBacktestConfig, GpuBacktestEvaluator, WindowMetrics,
 };
 use ml_dqn::evaluation::engine::EvaluationEngine;
+use ml_dqn::evaluation::metrics::OHLCVBarF32;
+use ml_core::common::action::FactoredAction;
 use candle_core::{Device, Tensor, DType};
 
 /// Generate deterministic synthetic price data (random walk with drift).
@@ -2184,15 +2186,22 @@ async fn test_gpu_vs_cpu_backtest_agreement() {
     ).expect("GPU evaluation");
 
     // 3. Run CPU backtest with same always-Flat action
+    use ml_core::common::action::{ExposureLevel, OrderType, Urgency};
+    let flat_action = FactoredAction::new(ExposureLevel::Flat, OrderType::Market, Urgency::Normal);
+
     let mut cpu_engines: Vec<EvaluationEngine> = vec![
-        EvaluationEngine::new(100_000.0),
-        EvaluationEngine::new(100_000.0),
+        EvaluationEngine::new_with_kelly(100_000.0, 1.0),
+        EvaluationEngine::new_with_kelly(100_000.0, 1.0),
     ];
-    let all_closes = [&closes_1, &closes_2];
-    for (engine, closes) in cpu_engines.iter_mut().zip(all_closes.iter()) {
-        for close in closes.iter() {
-            // Action 2 = Flat (no trading)
-            engine.process_bar_factored(*close, 2, 0.1, 0.0001);
+    let all_prices = [&prices_1, &prices_2];
+    for (engine, window_prices) in cpu_engines.iter_mut().zip(all_prices.iter()) {
+        for (bar_idx, ohlc) in window_prices.iter().enumerate() {
+            let bar = OHLCVBarF32 {
+                timestamp: bar_idx as i64,
+                open: ohlc[0], high: ohlc[1], low: ohlc[2], close: ohlc[3],
+                volume: 1000.0,
+            };
+            engine.process_bar_factored(bar_idx, &bar, &flat_action);
         }
     }