DARC Supporting Information - Publication Contents Manifest

This package is supplied in response to the production request for the complete Supporting Information.

Requested item -> supplied file(s)
1. Reproducibility code
   - reproduce_exact_benchmark.py
   - model_spec_current.py
2. Expected-reward matrix
   - expected_reward_matrix.npy
3. Transition tensor
   - transition_tensor.npy
4. Learned policy arrays
   - policy_darc.npy
   - policy_oracle.npy
   - policy_greedy.npy
   - policy_ucb.npy
   - policy_heuristic.npy
5. Ensemble Q-table
   - q_table_81state_ensemble.npy
6. Validation outputs
   - q_learning_seed_validation.csv
   - representative_multiseed_results.csv
   - dynamic_workload_benchmark.csv
   - exact_discounted_policy_benchmark.csv
   - reproduced_exact_discounted_policy_benchmark.csv
   - learned_policy_audit.csv
7. Plot-data files
   - p95_rtt_vs_hops.csv
   - goodput_vs_loss.csv
   - ablation.csv
   - message_size_sensitivity.csv
   - pareto_summary.csv
   - weight_sensitivity.csv
8. Model card
   - DARC_model_card_current.json
9. Empirical-calibration template
   - empirical_calibration_template.csv
10. Package guide
   - README_SUPPLEMENTARY.txt

The high-resolution manuscript figures are supplied separately in Figures.zip (Figures 1-10 plus the graphical abstract).

Scope note: The empirical calibration template is intentionally blank because the article does not claim new empirical RF calibration without raw timestamped RTT/loss traces.
