We argue that efficient agentic reasoning benefits from decomposing deliberation into three interacting systems: reactive execution (System I) for fine-grained reasoning and direct action; simulative reasoning (System II) that predicts consequences of proposed actions through a world model; and self-regulation (System III) that decides when and how deeply to plan through a learned configurator.
See the
GitHub repository for setup and inference instructions.
SR²AM-v0.1-8B sits above the size-vs-accuracy trendline in (a). The full benchmark breakdown is in the
paper.
1@article{deng2026sr2am,
2 title={Efficient Agentic Reasoning Through Self-Regulated Simulative Planning},
3 author={Deng, Mingkai and Hou, Jinyu and Neves, Lara Sá and
4 Pimpalkhute, Varad and Killian, Taylor W. and
5 Liu, Zhengzhong and Xing, Eric P.},
6 journal={arXiv preprint arXiv:2605.22138},
7 year={2026}
8}
