About xdrl#

xdrl lets TDHook operate on TorchRL modules and on the networks inside TorchRL losses. It selects the actor, critic, value, Q-value, mixer, or target parameters from the original TorchRL object; it does not describe the RL system again.

Literature#

The references below cover the foundations cited by XDRL’s public API and the papers targeted by its reproduction notebooks. Each reproduction notebook separately records the exact paper, reference-code revision, asset availability, and limits of the evidence it produces.

References#

[1]

Albert Bou, Matteo Bettini, Sebastian Dittert, Vikash Kumar, Shagun Sodhani, Xiaomeng Yang, Gianni De Fabritiis, and Vincent Moens. Torchrl: a data-driven decision-making library for pytorch. 2023. URL: https://arxiv.org/abs/2306.00577, arXiv:2306.00577.

[2]

Yoann Poupart. Tdhook: a lightweight framework for interpretability. 2025. URL: https://arxiv.org/abs/2509.25475, arXiv:2509.25475.

[3]

Anna Soligo, Pietro Ferraro, and David Boyle. Inducing, detecting and characterising neural modules: a pipeline for functional interpretability in reinforcement learning. In Proceedings of the 42nd International Conference on Machine Learning, volume 267 of Proceedings of Machine Learning Research, 56122–56147. PMLR, 2025. URL: https://proceedings.mlr.press/v267/soligo25a.html.

[4]

Thomas Bush, Stephen Chung, Usman Anwar, Adrià Garriga-Alonso, and David Krueger. Interpreting emergent planning in model-free reinforcement learning. In International Conference on Learning Representations. 2025. URL: https://proceedings.iclr.cc/paper_files/paper/2025/hash/cc4d9cfc45325e460b455a820d5f212c-Abstract-Conference.html.

[5]

Renos Zabounidis, Joseph Campbell, Simon Stepputtis, Dana Hughes, and Katia P. Sycara. Concept learning for interpretable multi-agent reinforcement learning. In Proceedings of The 6th Conference on Robot Learning, volume 205 of Proceedings of Machine Learning Research, 1828–1837. PMLR, 2023. URL: https://proceedings.mlr.press/v205/zabounidis23a.html.

[6]

Ulisse Mini, Peli Grietzer, Mrinank Sharma, Austin Meek, Monte MacDiarmid, and Alexander Matt Turner. Understanding and controlling a maze-solving policy network. 2023. URL: https://arxiv.org/abs/2310.08043, arXiv:2310.08043.

[7]

Zichuan Liu, Yuanyang Zhu, and Chunlin Chen. NA2Q: neural attention additive model for interpretable multi-agent q-learning. In Proceedings of the 40th International Conference on Machine Learning, volume 202 of Proceedings of Machine Learning Research, 22539–22558. PMLR, 2023. URL: https://proceedings.mlr.press/v202/liu23be.html.