SteelAgent: An LLM-Orchestrated System for Physics-Informed Steel Property Prediction and Generalization Auditing

Volkov, Aleksandr; Sultimov, Roman; Kuzin, Mikhail; Maximov, Yury

doi:10.24963/ijcai.2026/1000

SteelAgent: An LLM-Orchestrated System for Physics-Informed Steel Property Prediction and Generalization Auditing

Aleksandr Volkov, Roman Sultimov, Mikhail Kuzin, Yury Maximov

Download BibTeX Cite

Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence. Demo Track. Pages 8547-8550.

https://doi.org/10.24963/ijcai.2026/1000

Keywords

AI: Agent-based and Multi-agent Systems
AI: Multidisciplinary Topics and Applications
AI: Natural Language Processing
AI: Knowledge Representation and Reasoning

Abstract

Machine learning models for steel property prediction routinely report high-quality metrics with R² > 0.85, yet these results rely on random splits that allow similar grades in both train and test sets. We present SteelAgent, an interactive system that exposes a critical generalization gap: the same models drop from R² > 0.85 to R² = 0.11 on unseen steel families, revealing more than 7 times higher quality degradation. Similarly, conformal prediction coverage degrades from 91% to 38% under distribution shift induced by holding out substantial data sources. SteelAgent combines physics-informed features grounded in classical metallurgy and interpretable models with conformal uncertainty quantification, and an LLM orchestrator that coordinates six domain-specific tools. The system supports property prediction with specification compliance checking, competitive steel comparison, and cost-aware inverse alloy design over 3,741 heat treatment records spanning 1,234 grades. All predictions are traceable through explicit tool calls, ensuring that all physical quantities are computed, not generated. We made the code and data freely accessible to the community.

Aleksandr Volkov, Roman Sultimov, Mikhail Kuzin, and Yury Maximov. 2026. SteelAgent: An LLM-Orchestrated System for Physics-Informed Steel Property Prediction and Generalization Auditing. In Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26. IJCAI, 8547–8550. https://doi.org/10.24963/ijcai.2026/1000

@inproceedings{ijcai2026p1000,
  title     = {SteelAgent: An LLM-Orchestrated System for Physics-Informed Steel Property Prediction and Generalization Auditing},
  author    = {Volkov, Aleksandr and Sultimov, Roman and Kuzin, Mikhail and Maximov, Yury},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on
               Artificial Intelligence, {IJCAI-26}},
  publisher = {International Joint Conferences on Artificial Intelligence Organization},
  editor    = {Diego Calvanese},
  pages     = {8547--8550},
  year      = {2026},
  month     = {8},
  note      = {Demo Track},
  doi       = {10.24963/ijcai.2026/1000},
  url       = {https://doi.org/10.24963/ijcai.2026/1000},
}

TY  - CPAPER
TI  - SteelAgent: An LLM-Orchestrated System for Physics-Informed Steel Property Prediction and Generalization Auditing
AU  - Volkov, Aleksandr
AU  - Sultimov, Roman
AU  - Kuzin, Mikhail
AU  - Maximov, Yury
T2  - Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence, IJCAI-26
PB  - International Joint Conferences on Artificial Intelligence Organization
PY  - 2026
SP  - 8547
EP  - 8550
DO  - 10.24963/ijcai.2026/1000
UR  - https://doi.org/10.24963/ijcai.2026/1000
ER  -