Add final report, plots, experiment guide, and complete NOTE.md

All experiments complete:
- Toy LQ: credit bridge matches state bridge (~0.94 costate cosine)
- CIFAR-10: credit bridge (29.6%) comparable to DFA (30.0%), both beat state bridge (18.5%)
- State bridge confirms core hypothesis: perfect state prediction != useful credit
- Terminal gradient matching is essential for credit bridge

1 files changed, 1 insertions, 1 deletions

diff --git a/experiments/plot_toy_final.py b/experiments/plot_toy_final.py
index 2f7c109..e4e1a68 100644
--- a/experiments/plot_toy_final.py
+++ b/experiments/plot_toy_final.py
@@ -13,7 +13,7 @@ os.makedirs(output_dir, exist_ok=True)
 seeds = [42, 123, 456]
 all_data = []
 for seed in seeds:
-    path = f'results/toy_lq/toy_lq_v2_seed{seed}_lam0.1_sig0.1_tgw1.0_fm0.0.json'
+    path = f'results/toy_lq_frozen/toy_lq_v2_seed{seed}_lam0.1_sig0.1_tgw1.0_fm0.0.json'
     if os.path.exists(path):
         with open(path) as f:
             all_data.append(json.load(f))