Publications:
2.5-D Decomposition for LLM-Based Spatial Construction
Abstract: Autonomous systems that build structures from natural-language instructions need reliable spatial reasoning, yet large language models (LLMs) make systematic coordinate errors when generating three-dimensional block placements. We present a neuro-symbolic pipeline based on 2.5-D decomposition: the LLM plans in the two-dimensional horizontal plane while a deterministic executor computes all vertical placements from column occupancy, eliminating an entire class of errors. On the Build What I Mean benchmark (160 rounds), GPT-4o-mini with this pipeline achieves 94.6% mean structural accuracy across 12 independent runs, within 3.0 percentage points of the 97.6% ceiling imposed by architect-agent errors that no builder-side improvement can address. This outperforms both GPT-4o at 90.3% and the best competing system at 76.3%. A controlled ablation confirms that 2.5-D decomposition is the dominant contributor, accounting for 28.7 percentage points of accuracy. The pipeline transfers directly to edge hardware: Nemotron-3 120B on an NVIDIA Jetson Thor AGX achieves 96.0% mean structural accuracy with the identical pipeline, slightly exceeding the cloud result. Expanding the system prompt by four targeted examples to exceed the model's 8,320-token prefix cache page size, combined with low-effort reasoning, reduces mean per-request latency by 3X to 19.7 seconds at 95.6% accuracy. The underlying principle, removing deterministic dimensions from the LLM's output space, applies to any autonomous construction or assembly task where gravity or other physical constraints fix one or more degrees of freedom. A transfer experiment on 500 IGLU collaborative building tasks confirms the effect generalizes beyond the primary benchmark.
@INPROCEEDINGS{11674959,
author={Whitten, Paul and Chen, Li-Jen and Baddam, Sharath},
booktitle={NAECON 2026 - IEEE National Aerospace and Electronics Conference},
title={2.5-D Decomposition for LLM-Based Spatial Construction},
year={2026},
volume={},
number={},
pages={114-119},
keywords={Modeling;Large language models;Generative Pre-trained transformer;Planning;Accuracy;Cognition;Cognitive systems;Pipelines;Printing;Color;large language models;spatial reasoning;neurosymbolic systems;autonomous construction;2.5-D decomposition},
doi={10.1109/NAECON70028.2026.11674959}
}
Explainability Methods for Hardware Trojan Detection: A Systematic Comparison
@article{whitten2026explainability,
title={Explainability Methods for Hardware Trojan Detection: A Systematic Comparison},
author={Whitten, Paul and Wolff, Francis and Papachristou, Chris},
journal={Journal of Electronic Testing},
pages={1--21},
year={2026},
publisher={Springer}
}
Explainable AI Architectures: Methods, Applications, Examples, and Results
Video presentation - Dissertation source
@phdthesis{whitten2025explainable,
title={Explainable AI Architectures: Methods, Applications, Examples, and Results},
author={Whitten, Paul C},
year={2025},
school={Case Western Reserve University},
url={http://rave.ohiolink.edu/etdc/view?acc_num=case1743462769378071}
}
An AI Architecture with the Capability to Classify and Explain Hardware Trojans
Abstract: Hardware trojan detection methods, based on machine learning (ML) techniques, mainly identify suspected circuits but lack the ability to explain how the decision was arrived at. An explainable methodology and architecture is introduced based on the existing hardware trojan detection features. Results are provided for explaining digital hardware trojans within a netlist using trust-hub trojan benchmarks.
@INPROCEEDINGS{10670684,
author={Whitten, Paul and Wolff, Francis and Papachristou, Chris},
booktitle={NAECON 2024 - IEEE National Aerospace and Electronics Conference},
title={An AI Architecture with the Capability to Classify and Explain Hardware Trojans},
year={2024},
volume={},
number={},
pages={349-354},
keywords={Training;Support vector machines;Measurement;Accuracy;Costs;Aerospace electronics;Feature extraction;Hardware Trojan;Explainable Artificial Intel-ligence Architecture;XAI;CIA;Machine Learning;Support Vector Machine},
doi={10.1109/NAECON61878.2024.10670684}
}
An AI Architecture with the Capability to Explain Recognition Results
Abstract: Explainability is needed to establish confidence in machine learning results. Some explainable methods take a post hoc approach to explain the weights of machine learning models, others highlight areas of the input contributing to decisions. These methods do not adequately explain decisions, in plain terms. Explainable property-based systems have been shown to provide explanations in plain terms, however, they have not performed as well as leading unexplainable machine learning methods. This research focuses on the importance of metrics to explainability and contributes two methods yielding performance gains. The first method introduces a combination of explainable and unexplainable flows, proposing a metric to characterize explainability of a decision. The second method compares classic metrics for estimating the effectiveness of neural networks in the system, posing a new metric as the leading performer. Results from the new methods and examples from handwritten datasets are presented.
@INPROCEEDINGS{10586116,
author={Whitten, Paul and Wolff, Francis and Papachristou, Chris},
booktitle={2024 IEEE 3rd International Conference on Computing and Machine Intelligence (ICMI)},
title={An AI Architecture with the Capability to Explain Recognition Results},
year={2024},
volume={},
number={},
pages={1-6},
keywords={Measurement;Accuracy;Computational modeling;Neural networks;Machine learning;Computer architecture;Performance gain},
doi={10.1109/ICMI60790.2024.10586116}
}
Explainable Neural Network Recognition of Handwritten Characters
Abstract: The prevalence of Artificial Intelligence (AI), in many aspects of life, has brought about an increasing need for explainability of AI solutions. Previous work has been posed to interpret the weights of Neural Networks and provide visual hints of explainability. This work introduces a partitioning approach, constructing an explainable architecture that recognizes handwritten characters. The accuracy of the architecture is improved by introducing unexplainable components and a metric to characterize explainability. Ambiguous and mislabeled samples in training data prove challenging. Techniques are applied to identify issues and prune the training set to improve explainability. Pruned samples from the training set are presented to the architecture in an attempt to resolve issues. Results and explainable rationale from pruned data is presented.
@INPROCEEDINGS{10099288,
author={Whitten, Paul and Wolff, Francis and Papachristou, Chris},
booktitle={2023 IEEE 13th Annual Computing and Communication Workshop and Conference (CCWC)},
title={Explainable Neural Network Recognition of Handwritten Characters},
year={2023},
volume={},
number={},
pages={0176-0182},
keywords={Training;Measurement;Handwriting recognition;Visualization;Training data;Computer architecture;Artificial neural networks;Explainable Artificial Intelligence;XAI;Neural Network;Machine Learning;Training Set Pruning},
doi={10.1109/CCWC57344.2023.10099288}
}
Explainable Artificial Intelligence Methodology for Handwritten Applications
Abstract: There has been explosive growth of practical AI in recent years. A major concern of current AI systems and compliance regulations is an inability to explain inferential decisions. This work explores an Explainable Artificial Intelligence (XAI) methodology that provides explanations for classification decisions. Experimental results using the MNIST handwritten digit database are provided with explainable conclusions.
@INPROCEEDINGS{9696413,
author={Whitten, Paul and Wolff, Francis and Papachristou, Chris},
booktitle={NAECON 2021 - IEEE National Aerospace and Electronics Conference},
title={Explainable Artificial Intelligence Methodology for Handwritten Applications},
year={2021},
volume={},
number={},
pages={277-282},
keywords={Databases;Conferences;Aerospace electronics;Regulation;Explosives;Artificial intelligence;explainable;artificial intelligence;machine learning;AI regulations},
doi={10.1109/NAECON49338.2021.9696413}
}