🧪 Test?View on arXiv
SteelAgent: LLM-Orchestrated, Physics-Informed Steel Property Prediction with Audited Generalization and Calibrated Uncertainty
Not provided in the abstract
machine-learningmaterial-scienceuncertainty-calibration
2611.06953
Builder Relevance
3h ago70%
Abstract
SteelAgent integrates metallurgical descriptors and machine-learning models with a language-model interface for enhanced steel property prediction and uncertainty reporting.
Reality Card
Core Claim
SteelAgent demonstrates significant variability in predictive performance based on the choice of held-out units, highlighting the importance of context in model evaluation.
Method / Result
Pooled out-of-fold $R^2$ dropped from 0.859 to 0.113 when steel families were held out.
Limitations
Calibration can deteriorate on held-out groups, affecting the reliability of predictions.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.