Agentic Physical AI toward a Domain-Specific Foundation Model for Energy Systems: A Case Study on Nuclear Reactor Control
Yoon Pyo Lee, Samrendra Roy, Kazuma Kobayashi, Sajedul Talukder, Diab Abueidda, Seid Koric, Souvik Chakraborty, Syed Bahauddin Alam
Why It Matters
What makes this one worth your time
This work is relevant for AI researchers and engineers interested in applying AI to safety-critical systems, offering insights into domain-specific model development and reliability in controlled simulations.
A domain-specific AI model for nuclear reactor control shows promise in simulation but lacks real-world validation.
Summary
The paper proposes a domain-specific foundation model for nuclear reactor control, using a compact language model trained on synthetic scenarios. It emphasizes policy optimization through physics-based simulator validation, achieving significant reliability gains in simulation. The model shows potential for integration into a broader safety architecture but does not yet address off-nominal conditions or uncertainty quantification.
Key contributions
- Development of a compact language model for nuclear reactor control.
- Demonstration of significant reliability gains in simulated scenarios.
- Proposal of a model as part of a verification and monitoring architecture.
Notable insights
- The model achieves significant variance reduction and reliability gains without reinforcement learning or reward engineering.
- Representations transfer across simulators without architectural changes, indicating potential for broader applicability.
Possible limitations
- Does not address off-nominal operation, sensor faults, or uncertainty quantification.
- Not stated in the abstract
Abstract
arXiv:2512.23292v5 Announce Type: replace Abstract: The prevailing paradigm in AI for physical systems: scaling general-purpose foundation models toward universal multimodal reasoning, confronts a barrier at the control interface. Frontier vision-language models achieve only 50-53% accuracy on basic quantitative physics tasks, behaving as approximate guessers that preserve semantic plausibility while violating physical constraints. Safety-critical control demands outcome-space guarantees over executed actions, not parameter-space imitation. Here we present a pathway toward domain-specific foundation models through compact language models operating as Agentic Physical AI: policy optimization driven by physics-based simulator validation rather than perceptual inference. We train a 360M-parameter model on synthetic nuclear reactor scenarios scaled from 10^3 to 10^5 examples. Scaling produces strong, regime-dependent reliability gains under nominal simulated conditions, with variance collapse of approximately 500x and elimination of >10% terminal-power excursions on the sampled distribution. Despite balanced exposure to four actuation families, the model concentrates 95% of runtime execution on a single-bank strategy, without reinforcement learning or reward engineering. Representations transfer across simulators without architectural change. We position the system as a candidate decision component within a verification, monitoring, and defense-in-depth architecture, not as a stand-alone safety solution: the demonstrated behavior speaks to closed-loop reliability on a single-step task in simulation and does not yet address off-nominal operation, sensor faults, or uncertainty quantification.