jibingyang111/RT-A3C-MLN-ResearchData

RT-A3C-MLN reported experimental data

This archive was organized from:

• main(12).tex — primary structured source.
• RT_A3C_MLN(10).pdf — compiled-paper cross-check.

Important scope note

The paper reports aggregated experimental results such as mean ± standard deviation, confidence intervals, and summary values. These reported aggregates have been separated into machine-readable CSV columns.

They are not seed-level, episode-level, window-level, or sample-level raw logs, because those underlying run logs are not embedded in the uploaded main.tex or PDF and cannot be reconstructed exactly from aggregate results.

Files 06_component_search_ablation.csv, 10_gate_tau_sensitivity.csv, 11_gate_K_sensitivity.csv, and 12_label_delay_robustness.csv contain the exact source values used to generate the corresponding current PDF figures. The exact point arrays for the actor-count, concept-box, memory-capacity, and full training-curve figures were not present in the uploaded sources. Their availability is documented in 22_experiment_figure_data_availability.csv.

Format conventions

• Encoding: UTF-8 with BOM (utf-8-sig) for compatibility with Excel and common data tools.
• Delimiter: comma.
• Decimal separator: period.
• *_mean and *_std split values reported as mean ± standard deviation.
• *_estimate, *_ci_low, and *_ci_high store central estimates and bracketed intervals.
• *_comparator stores inequality signs such as < or >.
• *_reported preserves the cleaned value as displayed in the paper.
• Missing or inapplicable values are blank.

Source hashes

• main(12).tex SHA-256: a98d445d8637590a25e25aa53bf59895a4165c8ae2c6f0240e6e42994ab9be45
• RT_A3C_MLN(10).pdf SHA-256: e54ab3804096ed949c611a729d2c6ea04bb0fa8ea0b0c6e726239aaf799a6313