Catalyzing Informed Residential Energy Retrofit Decisions via Domain-Specific LLM

Armin Yeganeh; Dong Zhao; Jianli Chen; Jiayu Zhou; Lei Shu; Sinem Mollaoglu

arxiv: 2602.20181 · v2 · submitted 2026-02-19 · 💻 cs.CY · cs.AI

Catalyzing Informed Residential Energy Retrofit Decisions via Domain-Specific LLM

Lei Shu , Dong Zhao , Jianli Chen , Armin Yeganeh , Sinem Mollaoglu , Jiayu Zhou This is my paper

Pith reviewed 2026-05-15 20:35 UTC · model grok-4.3

classification 💻 cs.CY cs.AI

keywords residential energy retrofitdomain-specific LLMenergy efficiencyphysics-based simulationLoRA fine-tuninghomeowner decision supportCO2 reductionpayback period

0 comments

The pith

A fine-tuned LLM trained on physics simulations of 536,416 homes recommends high-quality energy retrofits from basic natural-language descriptions alone.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper creates a specialized large language model that helps homeowners select energy retrofits without technical expertise. The model uses low-rank adaptation fine-tuning on a dataset drawn from physics-based simulations and economic calculations across over half a million U.S. residential building prototypes. It covers nine retrofit categories such as envelope upgrades, HVAC systems, and renewables. Validations show the model matches physics-grounded optima at top-3 rates of 98.9 percent for greatest CO2 reduction and 93.3 percent for shortest discounted payback period. Performance remains stable even when input descriptions are only 60 percent complete, lowering the barrier for non-experts to reach informed choices that could scale to larger energy and emission savings.

Core claim

The authors establish that a domain-specific LLM, fine-tuned via LoRA on a corpus of physics-grounded energy simulations and techno-economic data from 536,416 U.S. residential building prototypes, consistently identifies high-quality retrofit options using only homeowner-accessible natural-language inputs such as building age, size, and location, with top-3 hit rates reaching 98.9 percent for maximum CO2 reduction and 93.3 percent for shortest discounted payback year while remaining robust under partial input conditions.

What carries the argument

The domain-specific LLM created by LoRA fine-tuning on a massive corpus of physics-based energy simulations and techno-economic calculations from 536,416 U.S. residential building prototypes.

If this is right

Homeowners gain access to high-quality retrofit options without needing structured technical assessments.
The model supports scalable, parallelized decision-making at community and national levels.
Cumulative energy savings and emission reductions accelerate through widespread user-centered choices.
Recommendations remain reliable even when basic dwelling information is only partially available.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Integration with mobile apps could let users upload photos or utility bills to refine recommendations further.
The same simulation-grounded fine-tuning method could extend to related household decisions such as water efficiency or solar sizing.
Pilot deployments in real neighborhoods would test whether the model's outputs increase actual retrofit adoption rates.

Load-bearing premise

The 536,416 simulated building prototypes accurately represent real-world U.S. residential buildings and natural-language descriptions of basic attributes suffice for reliable retrofit recommendations without additional technical details.

What would settle it

Compare the model's retrofit recommendations on descriptions of actual occupied homes against detailed physics simulations or measured post-retrofit energy use and costs for the same homes.

read the original abstract

Residential energy retrofit initiation is often stalled by an expertise gap, where homeowners lack the technical literacy required for structured building energy assessments and are thereby trapped in low-information environments with fragmented sources. To bridge this gap, this study reports a domain-specific large language model (LLM) designed to catalyze informed decision-making based solely on homeowner-accessible, natural-language descriptions, e.g., building age, size, and location. The model is created using the parameter-efficient low-rank adaption (LoRA) fine-tuning approach on a massive corpus grounded in physics-based energy simulations and techno-economic calculations from 536,416 U.S. residential building prototypes. Nine major retrofit categories are evaluated, including envelope upgrades, HVAC systems, and renewable energy installations. Validations against physics-grounded benchmarks show that the LLM consistently identifies high-quality retrofit options, achieving top-3 hit rates of 98.9% for maximum CO2 reduction and 93.3% for the shortest discounted payback year. Moreover, the model exhibits strong robustness under incomplete input conditions, maintaining stable performance even when basic dwelling descriptions are only 60% partially specified. By significantly lowering the information activation energy for non-expert users while maintaining the scientific rigor, this physics-based AI model offers a scalable pathway for parallelized, user-centered decision making, accelerating cumulative energy savings and emission reductions across community and national scales.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The retrofit LLM performs well inside its simulation training set but lacks evidence for real buildings.

read the letter

The key takeaway is that this work fine-tunes an LLM with LoRA on a massive set of physics-based energy simulations covering 536,416 building prototypes. The goal is to let homeowners get retrofit recommendations just from basic natural language descriptions, without needing technical expertise. The reported results show the model hitting 98.9% top-3 accuracy for identifying options with maximum CO2 reduction and 93.3% for shortest discounted payback periods, based on the simulation benchmarks. It also stays reliable when inputs are only partially specified. What the paper does well is demonstrate a practical way to connect general language models to detailed building physics and economic models. By training on that volume of simulated data, they create a system that can handle queries in everyday language while staying tied to the underlying calculations. The handling of incomplete information is particularly useful, since real users often won't provide every detail. On the downside, the validation stays entirely within the simulated prototypes. The high hit rates are measured against benchmarks derived from the same data used in training, so they don't test how the model performs when faced with actual buildings that might differ in unmodeled ways. Things like specific construction variations, real occupancy behaviors, or local factors not captured in the prototypes could affect the recommendations. The paper would be stronger with some form of external validation or at least a clearer analysis of where the simulation assumptions might break down in practice. This kind of work is aimed at researchers and developers focused on using AI to support energy efficiency decisions. Someone looking at applications of LLMs in sustainability or building science would find the approach relevant for seeing how to ground models in simulation outputs. I think it deserves peer review. The idea is straightforward and the numbers are promising enough that referees should examine the full methods to assess the generalization claims.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces a domain-specific LLM fine-tuned via LoRA on physics-based energy simulations and techno-economic calculations from 536,416 U.S. residential building prototypes. Using only natural-language inputs (e.g., age, size, location, with robustness to 60% partial specification), the model recommends among nine retrofit categories and is validated against physics-grounded benchmarks, reporting top-3 hit rates of 98.9% for maximum CO2 reduction and 93.3% for shortest discounted payback period.

Significance. If the sim-to-real transfer holds, the work could lower the expertise barrier for homeowners and enable scalable, user-centered retrofit decisions that accelerate energy savings and emissions reductions. The scale of the simulation corpus and the parameter-efficient fine-tuning approach are notable strengths that ground the recommendations in physics rather than purely data-driven patterns.

major comments (2)

[Abstract] Abstract: the reported top-3 hit rates (98.9% for CO2 reduction, 93.3% for payback) are obtained on the identical set of 536,416 simulated prototypes used for LoRA fine-tuning; this closed-distribution evaluation leaves the central claim of reliable recommendations for real-world natural-language inputs untested, as real buildings introduce unmodeled variables (construction details, occupancy, etc.) not captured by basic attribute descriptions.
[Abstract] Validation protocol (referenced in abstract): no information is supplied on data partitioning, train/test splits, baseline comparisons (e.g., against rule-based or general-purpose LLMs), or explicit handling of simulation-to-reality gaps, all of which are load-bearing for interpreting the hit-rate numbers as evidence of generalization.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. The comments highlight important aspects of our evaluation protocol that require clarification and expansion. We address each major comment below and will revise the manuscript to strengthen the presentation of our methods and limitations.

read point-by-point responses

Referee: [Abstract] Abstract: the reported top-3 hit rates (98.9% for CO2 reduction, 93.3% for payback) are obtained on the identical set of 536,416 simulated prototypes used for LoRA fine-tuning; this closed-distribution evaluation leaves the central claim of reliable recommendations for real-world natural-language inputs untested, as real buildings introduce unmodeled variables (construction details, occupancy, etc.) not captured by basic attribute descriptions.

Authors: We agree that the reported metrics reflect in-distribution performance on the full set of 536,416 simulated prototypes. This design choice allows direct comparison against physics-grounded ground truth for every archetype, confirming that the LoRA-adapted model accurately reproduces the simulation engine's optimal retrofit recommendations when given the corresponding natural-language descriptions. The 60% partial-specification robustness tests further demonstrate practical utility under incomplete homeowner inputs. We acknowledge that this does not constitute a direct test of sim-to-real transfer. In the revised manuscript we will add an explicit limitations paragraph in the Discussion section that enumerates unmodeled real-world factors (occupancy schedules, detailed envelope construction, micro-climate effects) and outline planned follow-on studies that pair the model with field data from instrumented homes. revision: yes
Referee: [Abstract] Validation protocol (referenced in abstract): no information is supplied on data partitioning, train/test splits, baseline comparisons (e.g., against rule-based or general-purpose LLMs), or explicit handling of simulation-to-reality gaps, all of which are load-bearing for interpreting the hit-rate numbers as evidence of generalization.

Authors: We will expand the Methods section to document the validation protocol in full. Because the corpus consists of unique, exhaustively simulated archetypes rather than sampled real buildings, we trained and evaluated on the complete set to ensure coverage of all U.S. residential building types; we will state this rationale explicitly. We will also insert baseline comparisons: (1) a deterministic rule-based recommender that applies the same techno-economic criteria used to generate the ground truth, and (2) zero-shot and few-shot prompting of an unmodified general-purpose LLM. These additions will quantify the performance lift attributable to domain-specific LoRA fine-tuning. Finally, we will add a dedicated subsection on simulation-to-reality considerations, referencing the partial-input robustness results as preliminary evidence of tolerance to missing attributes and outlining the data-collection steps needed for future out-of-distribution validation. revision: yes

Circularity Check

0 steps flagged

No significant circularity in derivation chain

full rationale

The paper constructs a training corpus from 536,416 physics-based simulations, applies LoRA fine-tuning to map natural-language building descriptions to retrofit recommendations, and reports empirical top-3 hit rates against held-out physics benchmarks drawn from the same simulation framework. This constitutes standard supervised learning with external ground truth; no equations, fitted parameters, or self-citations reduce the hit-rate metrics to tautological definitions or inputs by construction. The derivation chain remains self-contained as an empirical ML performance result rather than a self-referential loop.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Central claim depends on the representativeness of the simulation corpus and the sufficiency of minimal natural-language inputs; no free parameters or invented entities are explicitly introduced beyond standard LLM fine-tuning.

axioms (1)

domain assumption The 536,416 U.S. residential building prototypes generated via physics-based energy simulations and techno-economic calculations accurately capture real-world building stock and retrofit performance.
Invoked as the training and validation foundation in the abstract.

pith-pipeline@v0.9.0 · 5558 in / 1180 out tokens · 28938 ms · 2026-05-15T20:35:22.855180+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

51 extracted references · 51 canonical work pages · 1 internal anchor

[1]

Miotto, M

Zhao, D., A.B. Miotto, M. Syal, and J. Chen, Framework for Benchmarking green building movement: A case of Brazil. Sustainable Cities and Society, 2019. 48: p. 101545

work page 2019
[2]

Das, and M

Bawaneh, K., S. Das, and M. Rasheduzzaman, Energy Consumption Analysis and Characterization of the Residential Sector in the US towards Sustainable Development. Energies, 2024. 17(11): p. 2789

work page 2024
[3]

Shu, and D

Xu, C., L. Shu, and D. Zhao, Optimizing Building Energy Use Reduction: Integrating HVAC Systems and Building Envelope through Sensitivity Analysis, in Computing in Civil Engineering 2024. p. 317–327

work page 2024
[4]

Yeganeh, and D

Shu, L., A. Yeganeh, and D. Zhao, Large Language Models for Building Energy Retrofit Decision-Making: Technical and Sociotechnical Evaluations. Buildings, 2025. 15(22): p. 4081

work page 2025
[5]

Shu, L. and D. Zhao, Techno-Economic Analysis of Building Energy Retrofits: Integrating Occupant Behavior Impacts, in Computing in Civil Engineering 2024. 2024. p. 305–316

work page 2024
[6]

Shu, L. and D. Zhao. Data-Driven Residence Energy Consumption Prediction Model Considering Water Use Data and Socio -Demographic Data . in Construction Research Congress 2024. 2023

work page 2024
[7]

Shu, L. and D. Zhao, A Scalable Computational Framework for Evaluating Residential Energy Retrofits Across Diverse Climates and Occupant Behaviors. Journal of Computing in Civil Engineering, 2026(Forthcoming)

work page 2026
[8]

Shu, L., T. Hong, K. Sun, and D. Zhao, Framework to select robust energy retrofit measures for residential communities. Energy and Buildings, 2025. 327: p. 115077

work page 2025
[9]

Cincinelli, A. and T. Martellini, Indoor air quality and health. International Journal of Environmental Research and Public Health, 2017. 14(11): p. 1286

work page 2017
[10]

McCoy, and J

Zhao, D., A. McCoy, and J. Du, An empirical study on the energy consumption in residential buildings after adopting green building standards. Procedia Engineering, 2016. 145: p. 766–773

work page 2016
[11]

Shu, L. and D. Zhao, Decision-making approach to urban energy retrofit —a comprehensive review. Buildings, 2023. 13(6): p. 1425

work page 2023
[12]

Rai, V . and S.A. Robinson, Effective information channels for reducing costs of environmentally-friendly technologies: evidence from residential PV markets. Environmental Research Letters, 2013. 8(1): p. 014044

work page 2013
[13]

Shu, L., D. Zhao, W. Zhang, H. Li, and T. Hong, IoT-based retrofit information diffusion in future smart communities. Energy and Buildings, 2025. 338: p. 115756

work page 2025
[14]

Kerr, N. and M. Winskel, Household investment in home energy retrofit: A review of the evidence on effective public policy design for privately owned homes. Renewable and Sustainable Energy Reviews, 2020. 123: p. 109778

work page 2020
[15]

Asadi, and J

Safari, M., S. Asadi, and J. Freihaut. Business development in small commercial building energy retrofit projects—a review on current industry practices. in Construction Research Congress 2020. 2020. American Society of Civil Engineers Reston, V A

work page 2020
[16]

Zaunbrecher, and M

Arning, K., B.S. Zaunbrecher, and M. Ziefle. The influence of intermediaries’ advice on energy-efficient retrofit decisions in private households. in Proceedings of the eceee. 2019

work page 2019
[17]

Meng, F., Z. Lu, X. Li, W. Han, J. Peng, X. Liu, and Z. Niu, Demand-side energy management reimagined: A comprehensive literature analysis leveraging large language models. Energy, 2024. 291: p. 130303

work page 2024
[18]

Zhang, J

Liu, M., L. Zhang, J. Chen, W.-A. Chen, Z. Yang, L.J. Lo, J. Wen, and Z. O’Neill. Large language models for building energy applications: Opportunities and challenges . in Building Simulation. 2025. Springer

work page 2025
[19]

Mo, and D

Shu, L., Y . Mo, and D. Zhao, Energy retrofits for smart and connected communities: Scopes and technologies. Renewable and Sustainable Energy Reviews, 2024. 199: p. 114510

work page 2024
[20]

Bianco, C

Ascione, F., N. Bianco, C. De Stasio, G.M. Mauro, and G.P. Vanoli, Artificial neural networks to predict energy performance and retrofit scenarios for any member of a building category: A novel approach. Energy, 2017. 118: p. 999–1017

work page 2017
[21]

Da Silva, C.H

Asadi, E., M.G. Da Silva, C.H. Antunes, L. Dias, and L. Glicksman, Multi-objective optimization for building retrofit: A model using genetic algorithm and artificial neural network and an application. Energy and buildings, 2014. 81: p. 444–456

work page 2014
[22]

He, and J

Zhan, J., W. He, and J. Huang, Dual-objective building retrofit optimization under competing priorities using Artificial Neural Network. Journal of Building Engineering,

work page
[23]

Mavromatidis, A

Thrampoulidis, E., G. Mavromatidis, A. Lucchi, and K. Orehounig, A machine learning- based surrogate model to approximate optimal building retrofit solutions. Applied Energy,

work page
[24]

Zhang, H., H. Feng, K. Hewage, and M. Arashpour, Artificial neural network for predicting building energy performance: a surrogate energy retrofits decision support framework. Buildings, 2022. 12(6): p. 829

work page 2022
[25]

Dzemyda, L

Kaklauskas, A., G. Dzemyda, L. Tupenaite, I. V oitau, O. Kurasova, J. Naimaviciene, Y . Rassokha, and L. Kanapeckiene, Artificial neural network-based decision support system for development of an energy-efficient built environment. Energies, 2018. 11(8): p. 1994

work page 2018
[26]

Gnekpe, and D

Nyawa, S., C. Gnekpe, and D. Tchuente, Transparent machine learning models for predicting decisions to undertake energy retrofits in residential buildings. Annals of Operations Research, 2023: p. 1–29

work page 2023
[27]

Choi, and R.K

Nutkiewicz, A., B. Choi, and R.K. Jain, Exploring the influence of urban context on building energy retrofit performance: A hybrid simulation and data -driven approach. Advances in Applied Energy, 2021. 3: p. 100038

work page 2021
[28]

Dai, and A

Deb, C., Z. Dai, and A. Schlueter, A machine learning-based framework for cost -optimal building retrofit. Applied energy, 2021. 294: p. 116990

work page 2021
[29]

Gomes, P

Araújo, G., R. Gomes, P. Ferrão, and M.G. Gomes, Optimizing building retrofit through data analytics: A study of multi-objective optimization and surrogate models derived from energy performance certificates. Energy and Built Environment, 2024. 5(6): p. 889–899

work page 2024
[30]

Shan, R., W. Lai, H. Tang, X. Leng, and W. Gu, Residential Building Renovation Considering Energy, Carbon Emissions, and Cost: An Approach Integrating Machine Learning and Evolutionary Generation. Applied Sciences, 2025. 15(4): p. 1830

work page 2025
[31]

Bano, M.H

Ali, U., S. Bano, M.H. Shamsi, D. Sood, C. Hoare, W. Zuo, N. Hewitt, and J. O'Donnell, Urban building energy performance prediction and retrofit analysis using data -driven machine learning approach. Energy and Buildings, 2024. 303: p. 113768

work page 2024
[32]

Zhong, and T

Li, K., W. Zhong, and T. Zhang, Improving building retrofit Decision -Making by integrating passive and BIPV techniques with ensemble model. Energy and Buildings, 2024. 323: p. 114727

work page 2024
[33]

Wang, B., H. Xi, W. Hou, and Y . Li, Low-carbon retrofit of rural dwellings in the dabie mountain region of China based on life-cycle assessment. Energy and Buildings, 2025: p. 115991

work page 2025
[34]

Luo, S., P.F. Yuan, M. Zhao, J. Yao, and F. Yang, Developing a Framework for Sustainable Retrofit of Residential Buildings Based on Ensemble Learning Algorithm: A Case Study of Shanghai. Building and Environment, 2025: p. 113311

work page 2025
[35]

Muzi, and Z

Piras, G., F. Muzi, and Z. Ziran, A Data -Driven Model for the Energy and Economic Assessment of Building Renovations. Applied Sciences, 2025. 15(14): p. 8117

work page 2025
[36]

Loftness, and E

Xu, Y ., V . Loftness, and E. Severnini, Using machine learning to predict retrofit effects for a commercial building portfolio. Energies, 2021. 14(14): p. 4334

work page 2021
[37]

Qiblawi, S

Markarian, E., S. Qiblawi, S. Krishnan, A. Divakaran, O. Ramalingam Rethnam, A. Thomas, and E. Azar, Informing building retrofits at low computational costs: A multi - objective optimisation using machine learning surrogates of building performance simulation models. Journal of Building Performance Simulation, 2024: p. 1–17

work page 2024
[38]

Zhang, L. and Z. Chen, Opportunities of applying Large Language Models in building energy sector. Renewable and Sustainable Energy Reviews, 2025. 214: p. 115558

work page 2025
[39]

Jiang, G., Z. Ma, L. Zhang, and J. Chen, EPlus-LLM: A large language model -based computing platform for automated building energy modeling. Applied Energy, 2024. 367: p. 123431

work page 2024
[40]

Xu, Y ., S. Zhu, J. Cai, J. Chen, and S. Li, A large language model-based platform for real- time building monitoring and occupant interaction. Journal of Building Engineering, 2025. 100: p. 111488

work page 2025
[41]

Choi, S. and S. Yoon, GPT-based data -driven urban building energy modeling (GPT - UBEM): Concept, methodology, and case studies. Energy and Buildings, 2024. 325: p. 115042

work page 2024
[42]

Prol Godoy, J

Hidalgo-Betanzos, J.M., I. Prol Godoy, J. Terés Zubiaga, R. Briones Llorente, and A. Martín Garin, Can ChatGPT AI Replace or Contribute to Experts’ Diagnosis for Renovation Measures Identification? Buildings, 2025. 15(3): p. 421

work page 2025
[43]

Darko, F

Chen, L., A. Darko, F. Zhang, A.P. Chan, and Q. Yang, Can large language models replace human experts? Effectiveness and limitations in building energy retrofit challenges assessment. Building and Environment, 2025. 276: p. 112891

work page 2025
[44]

National Laboratory of the Rockies ResStock Dataset 2024.2 . 2024; Available from: https://data.openei.org/s3_viewer?bucket=oedi-data-lake&prefix=nrel-pds-building- stock%2Fend-use-load-profiles-for-us-building- stock%2F2024%2Fresstock_tmy3_release_2%2F

work page 2024
[45]

National Residential Efficiency Measures Database

NLR. National Residential Efficiency Measures Database . 2018 [cited 2024; Available from: https://remdb.nrel.gov/

work page 2018
[46]

2021, National Renewable Energy Laboratory

Bianchi, TMY3 Weather Data for ComStock and ResStock , Fontanini, Editor. 2021, National Renewable Energy Laboratory

work page 2021
[47]

Yang, A., A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, and C. Lv, Qwen3 technical report. arXiv preprint arXiv:2505.09388, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[48]

Feng, Z., Y . Xie, J. Yang, W. Hou, and Z. Li. A Survey of Low-Rank Adaptation Techniques. in 2025 8th International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE). 2025. IEEE

work page 2025
[49]

Goldman-Wetzler, E

Hayford, J., J. Goldman-Wetzler, E. Wang, and L. Lu, Speeding up and reducing memory usage for scientific machine learning via mixed precision. Computer Methods in Applied Mechanics and Engineering, 2024. 428: p. 117093

work page 2024
[50]

Comesana, C

Li, H., A. Comesana, C. Weyandt, and T. Hong, A RAG Data Pipeline Transforming Heterogeneous Data into AI -Ready Format for Autonomous Building Performance Discovery. Advances in Applied Energy, 2025: p. 100261

work page 2025
[51]

Xu, C., L. Shu, A. Dao, and Y . Cui, Multimodal generative AI for automated pavement condition assessment: Benchmarking model performance. PLoS One, 2026. 21(1)

work page 2026

[1] [1]

Miotto, M

Zhao, D., A.B. Miotto, M. Syal, and J. Chen, Framework for Benchmarking green building movement: A case of Brazil. Sustainable Cities and Society, 2019. 48: p. 101545

work page 2019

[2] [2]

Das, and M

Bawaneh, K., S. Das, and M. Rasheduzzaman, Energy Consumption Analysis and Characterization of the Residential Sector in the US towards Sustainable Development. Energies, 2024. 17(11): p. 2789

work page 2024

[3] [3]

Shu, and D

Xu, C., L. Shu, and D. Zhao, Optimizing Building Energy Use Reduction: Integrating HVAC Systems and Building Envelope through Sensitivity Analysis, in Computing in Civil Engineering 2024. p. 317–327

work page 2024

[4] [4]

Yeganeh, and D

Shu, L., A. Yeganeh, and D. Zhao, Large Language Models for Building Energy Retrofit Decision-Making: Technical and Sociotechnical Evaluations. Buildings, 2025. 15(22): p. 4081

work page 2025

[5] [5]

Shu, L. and D. Zhao, Techno-Economic Analysis of Building Energy Retrofits: Integrating Occupant Behavior Impacts, in Computing in Civil Engineering 2024. 2024. p. 305–316

work page 2024

[6] [6]

Shu, L. and D. Zhao. Data-Driven Residence Energy Consumption Prediction Model Considering Water Use Data and Socio -Demographic Data . in Construction Research Congress 2024. 2023

work page 2024

[7] [7]

Shu, L. and D. Zhao, A Scalable Computational Framework for Evaluating Residential Energy Retrofits Across Diverse Climates and Occupant Behaviors. Journal of Computing in Civil Engineering, 2026(Forthcoming)

work page 2026

[8] [8]

Shu, L., T. Hong, K. Sun, and D. Zhao, Framework to select robust energy retrofit measures for residential communities. Energy and Buildings, 2025. 327: p. 115077

work page 2025

[9] [9]

Cincinelli, A. and T. Martellini, Indoor air quality and health. International Journal of Environmental Research and Public Health, 2017. 14(11): p. 1286

work page 2017

[10] [10]

McCoy, and J

Zhao, D., A. McCoy, and J. Du, An empirical study on the energy consumption in residential buildings after adopting green building standards. Procedia Engineering, 2016. 145: p. 766–773

work page 2016

[11] [11]

Shu, L. and D. Zhao, Decision-making approach to urban energy retrofit —a comprehensive review. Buildings, 2023. 13(6): p. 1425

work page 2023

[12] [12]

Rai, V . and S.A. Robinson, Effective information channels for reducing costs of environmentally-friendly technologies: evidence from residential PV markets. Environmental Research Letters, 2013. 8(1): p. 014044

work page 2013

[13] [13]

Shu, L., D. Zhao, W. Zhang, H. Li, and T. Hong, IoT-based retrofit information diffusion in future smart communities. Energy and Buildings, 2025. 338: p. 115756

work page 2025

[14] [14]

Kerr, N. and M. Winskel, Household investment in home energy retrofit: A review of the evidence on effective public policy design for privately owned homes. Renewable and Sustainable Energy Reviews, 2020. 123: p. 109778

work page 2020

[15] [15]

Asadi, and J

Safari, M., S. Asadi, and J. Freihaut. Business development in small commercial building energy retrofit projects—a review on current industry practices. in Construction Research Congress 2020. 2020. American Society of Civil Engineers Reston, V A

work page 2020

[16] [16]

Zaunbrecher, and M

Arning, K., B.S. Zaunbrecher, and M. Ziefle. The influence of intermediaries’ advice on energy-efficient retrofit decisions in private households. in Proceedings of the eceee. 2019

work page 2019

[17] [17]

Meng, F., Z. Lu, X. Li, W. Han, J. Peng, X. Liu, and Z. Niu, Demand-side energy management reimagined: A comprehensive literature analysis leveraging large language models. Energy, 2024. 291: p. 130303

work page 2024

[18] [18]

Zhang, J

Liu, M., L. Zhang, J. Chen, W.-A. Chen, Z. Yang, L.J. Lo, J. Wen, and Z. O’Neill. Large language models for building energy applications: Opportunities and challenges . in Building Simulation. 2025. Springer

work page 2025

[19] [19]

Mo, and D

Shu, L., Y . Mo, and D. Zhao, Energy retrofits for smart and connected communities: Scopes and technologies. Renewable and Sustainable Energy Reviews, 2024. 199: p. 114510

work page 2024

[20] [20]

Bianco, C

Ascione, F., N. Bianco, C. De Stasio, G.M. Mauro, and G.P. Vanoli, Artificial neural networks to predict energy performance and retrofit scenarios for any member of a building category: A novel approach. Energy, 2017. 118: p. 999–1017

work page 2017

[21] [21]

Da Silva, C.H

Asadi, E., M.G. Da Silva, C.H. Antunes, L. Dias, and L. Glicksman, Multi-objective optimization for building retrofit: A model using genetic algorithm and artificial neural network and an application. Energy and buildings, 2014. 81: p. 444–456

work page 2014

[22] [22]

He, and J

Zhan, J., W. He, and J. Huang, Dual-objective building retrofit optimization under competing priorities using Artificial Neural Network. Journal of Building Engineering,

work page

[23] [23]

Mavromatidis, A

Thrampoulidis, E., G. Mavromatidis, A. Lucchi, and K. Orehounig, A machine learning- based surrogate model to approximate optimal building retrofit solutions. Applied Energy,

work page

[24] [24]

Zhang, H., H. Feng, K. Hewage, and M. Arashpour, Artificial neural network for predicting building energy performance: a surrogate energy retrofits decision support framework. Buildings, 2022. 12(6): p. 829

work page 2022

[25] [25]

Dzemyda, L

Kaklauskas, A., G. Dzemyda, L. Tupenaite, I. V oitau, O. Kurasova, J. Naimaviciene, Y . Rassokha, and L. Kanapeckiene, Artificial neural network-based decision support system for development of an energy-efficient built environment. Energies, 2018. 11(8): p. 1994

work page 2018

[26] [26]

Gnekpe, and D

Nyawa, S., C. Gnekpe, and D. Tchuente, Transparent machine learning models for predicting decisions to undertake energy retrofits in residential buildings. Annals of Operations Research, 2023: p. 1–29

work page 2023

[27] [27]

Choi, and R.K

Nutkiewicz, A., B. Choi, and R.K. Jain, Exploring the influence of urban context on building energy retrofit performance: A hybrid simulation and data -driven approach. Advances in Applied Energy, 2021. 3: p. 100038

work page 2021

[28] [28]

Dai, and A

Deb, C., Z. Dai, and A. Schlueter, A machine learning-based framework for cost -optimal building retrofit. Applied energy, 2021. 294: p. 116990

work page 2021

[29] [29]

Gomes, P

Araújo, G., R. Gomes, P. Ferrão, and M.G. Gomes, Optimizing building retrofit through data analytics: A study of multi-objective optimization and surrogate models derived from energy performance certificates. Energy and Built Environment, 2024. 5(6): p. 889–899

work page 2024

[30] [30]

Shan, R., W. Lai, H. Tang, X. Leng, and W. Gu, Residential Building Renovation Considering Energy, Carbon Emissions, and Cost: An Approach Integrating Machine Learning and Evolutionary Generation. Applied Sciences, 2025. 15(4): p. 1830

work page 2025

[31] [31]

Bano, M.H

Ali, U., S. Bano, M.H. Shamsi, D. Sood, C. Hoare, W. Zuo, N. Hewitt, and J. O'Donnell, Urban building energy performance prediction and retrofit analysis using data -driven machine learning approach. Energy and Buildings, 2024. 303: p. 113768

work page 2024

[32] [32]

Zhong, and T

Li, K., W. Zhong, and T. Zhang, Improving building retrofit Decision -Making by integrating passive and BIPV techniques with ensemble model. Energy and Buildings, 2024. 323: p. 114727

work page 2024

[33] [33]

Wang, B., H. Xi, W. Hou, and Y . Li, Low-carbon retrofit of rural dwellings in the dabie mountain region of China based on life-cycle assessment. Energy and Buildings, 2025: p. 115991

work page 2025

[34] [34]

Luo, S., P.F. Yuan, M. Zhao, J. Yao, and F. Yang, Developing a Framework for Sustainable Retrofit of Residential Buildings Based on Ensemble Learning Algorithm: A Case Study of Shanghai. Building and Environment, 2025: p. 113311

work page 2025

[35] [35]

Muzi, and Z

Piras, G., F. Muzi, and Z. Ziran, A Data -Driven Model for the Energy and Economic Assessment of Building Renovations. Applied Sciences, 2025. 15(14): p. 8117

work page 2025

[36] [36]

Loftness, and E

Xu, Y ., V . Loftness, and E. Severnini, Using machine learning to predict retrofit effects for a commercial building portfolio. Energies, 2021. 14(14): p. 4334

work page 2021

[37] [37]

Qiblawi, S

Markarian, E., S. Qiblawi, S. Krishnan, A. Divakaran, O. Ramalingam Rethnam, A. Thomas, and E. Azar, Informing building retrofits at low computational costs: A multi - objective optimisation using machine learning surrogates of building performance simulation models. Journal of Building Performance Simulation, 2024: p. 1–17

work page 2024

[38] [38]

Zhang, L. and Z. Chen, Opportunities of applying Large Language Models in building energy sector. Renewable and Sustainable Energy Reviews, 2025. 214: p. 115558

work page 2025

[39] [39]

Jiang, G., Z. Ma, L. Zhang, and J. Chen, EPlus-LLM: A large language model -based computing platform for automated building energy modeling. Applied Energy, 2024. 367: p. 123431

work page 2024

[40] [40]

Xu, Y ., S. Zhu, J. Cai, J. Chen, and S. Li, A large language model-based platform for real- time building monitoring and occupant interaction. Journal of Building Engineering, 2025. 100: p. 111488

work page 2025

[41] [41]

Choi, S. and S. Yoon, GPT-based data -driven urban building energy modeling (GPT - UBEM): Concept, methodology, and case studies. Energy and Buildings, 2024. 325: p. 115042

work page 2024

[42] [42]

Prol Godoy, J

Hidalgo-Betanzos, J.M., I. Prol Godoy, J. Terés Zubiaga, R. Briones Llorente, and A. Martín Garin, Can ChatGPT AI Replace or Contribute to Experts’ Diagnosis for Renovation Measures Identification? Buildings, 2025. 15(3): p. 421

work page 2025

[43] [43]

Darko, F

Chen, L., A. Darko, F. Zhang, A.P. Chan, and Q. Yang, Can large language models replace human experts? Effectiveness and limitations in building energy retrofit challenges assessment. Building and Environment, 2025. 276: p. 112891

work page 2025

[44] [44]

National Laboratory of the Rockies ResStock Dataset 2024.2 . 2024; Available from: https://data.openei.org/s3_viewer?bucket=oedi-data-lake&prefix=nrel-pds-building- stock%2Fend-use-load-profiles-for-us-building- stock%2F2024%2Fresstock_tmy3_release_2%2F

work page 2024

[45] [45]

National Residential Efficiency Measures Database

NLR. National Residential Efficiency Measures Database . 2018 [cited 2024; Available from: https://remdb.nrel.gov/

work page 2018

[46] [46]

2021, National Renewable Energy Laboratory

Bianchi, TMY3 Weather Data for ComStock and ResStock , Fontanini, Editor. 2021, National Renewable Energy Laboratory

work page 2021

[47] [47]

Yang, A., A. Li, B. Yang, B. Zhang, B. Hui, B. Zheng, B. Yu, C. Gao, C. Huang, and C. Lv, Qwen3 technical report. arXiv preprint arXiv:2505.09388, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025

[48] [48]

Feng, Z., Y . Xie, J. Yang, W. Hou, and Z. Li. A Survey of Low-Rank Adaptation Techniques. in 2025 8th International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE). 2025. IEEE

work page 2025

[49] [49]

Goldman-Wetzler, E

Hayford, J., J. Goldman-Wetzler, E. Wang, and L. Lu, Speeding up and reducing memory usage for scientific machine learning via mixed precision. Computer Methods in Applied Mechanics and Engineering, 2024. 428: p. 117093

work page 2024

[50] [50]

Comesana, C

Li, H., A. Comesana, C. Weyandt, and T. Hong, A RAG Data Pipeline Transforming Heterogeneous Data into AI -Ready Format for Autonomous Building Performance Discovery. Advances in Applied Energy, 2025: p. 100261

work page 2025

[51] [51]

Xu, C., L. Shu, A. Dao, and Y . Cui, Multimodal generative AI for automated pavement condition assessment: Benchmarking model performance. PLoS One, 2026. 21(1)

work page 2026