Search NASA⌕ Search

SEARCH · Search NASA

Results for “LLMs”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 73 records · Page 4
All sourcesNASA NTRSDOE OSTI
NASA NTRSDate not supplied

Harnessing Large Language Models for Scientific Endeavors

The rapid proliferation of Large Language Models (LLMs) such as GPT, Bard, and Llama has revolutionized various sectors, including the scientific community. These models, with their potential to automate and augment tasks, are increasingly being recognized as both a valuable asset and a potential challenge in the realm of scientific research and data management. However, the current LLMs, primarily trained on general corpora, exhibit a limited understanding of scientific concepts and terminologies due to the lack of scientific corpus in their training data. Recognizing this gap, several groups are now advocating for the development of LLMs specifically tailored for scientific applications. A notable initiative in this direction is the Large Language Model effort initiated by NASA's CSDO. This endeavor aims to align LLM efforts across NASA’s Science Mission Directorate, develop a science-specific corpus and validation test set for model training, and create an encoder-only model for various downstream tasks. Moreover, the initiative also plans to develop a decoder-only model to explore the potential benefits and risks associated with a generative LLM for science. Lastly, the project aims to create a science evaluation suite, encompassing various categories of downstream scientific tasks, to serve as a benchmark for assessing the value of any LLM for future use. This presentation will provide an overview and current status of this ongoing initiative, highlighting its potential to reshape the use of LLMs in the scientific domain.

Rahul Ramachandran↗
DOE OSTI2025

Automation of Vulnerability and Patch Management: Information Extraction, Association, and Optimization

Vulnerability and patch management is an integral part of a robust cybersecurity program, yet it grows increasingly complex due to the sheer amount of data that must be analyzed. Particularly in Operational Technology (OT) environments, analysis must be done manually because of the lack of automated solutions. Additionally, there are many steps in this process, from the initial discovery of the vulnerability to the implementation of its remediation, and each step in the process requires different data in order to be performed effectively. In this work, we provide approaches and strategies to assist operators in industrial or OT environments throughout the vulnerability management cycle. Security advisories provide key information about mitigation strategies, or actions that can be taken when a patch is unavailable or cannot be installed. Details of these strategies are not shared in public vulnerability databases and must be found manually. We approach this problem by designing a solution to automatically identify that information within vendor security advisories and retrieve it for operator use. We start with an approach that requires domain-specific knowledge of certain frequently-seen reference websites. Next, an approach that can work on an arbitrary website but relies on certain keywords. Finally, an approach that uses Natural Language Processing (NLP) methods and does not require specific knowledge or keywords. Each of these approaches is more general than its predecessor; we demonstrate high accuracy for all approaches Advisories also often contain details of affected products in non-standard or natural language formats. While this information can be easily understood when read by an operator, the non-standard format acts as a barrier to effective automation. We provide an approach for the first step in this process: identifying vendors in security advisories and mapping them to a standard framework for representing digital assets and software products. We evaluate five established string similarity algorithms, plus one of our own design that combines string similarity and information theory, on the task of mapping vendors to their corresponding entries in the Common Platform Enumeration (CPE) repository. Our results show that our proposed metric outperforms all others. Due to the constraints on time, finances, and personnel for organizations, Large Language Models (LLMs) may seem like attractive opportunities for security operators to speed up information gathering; however, it is still not clear whether LLMs can handle vulnerability management tasks well. To answer this question, we perform an empirical study of LLMs’ ability to provide consistent, accurate information about vulnerabilities in order to guide organizations in their adoption of LLMs. We observe poor performance for all models tested, suggesting that these models are not well-suited to the consistent retrieval of accurate vulnerability information. Finally, once vulnerabilities have been identified and any additional information has been obtained, operators must decide which remediation actions to implement based on their available resources. This already-complex problem becomes even more so when we consider that a vulnerability may have multiple avenues for remediation. We formulate this scenario as two knapsack problems and provide solutions, which we then compare against several existing strategies for vulnerability prioritization seen in real operational environments.

McClanahan, Kylie↗
DOE OSTI2025

Can Large Language Models Understand Intermediate Representations?

Intermediate Representations (IRs) are essential in compiler design and program analysis, yet their comprehension by Large Language Models (LLMs) remains underexplored. This paper presents a pioneering empirical study to investigate the capabilities of LLMs, including GPT-4, GPT-3, Gemma 2, LLaMA 3.1, and Code Llama, in understanding IRs. We analyze their performance across four tasks: Control Flow Graph (CFG) reconstruction, decompilation, code summarization, and execution reasoning. Our results indicate that while LLMs demonstrate competence in parsing IR syntax and recognizing high-level structures, they struggle with control flow reasoning, execution semantics, and loop handling. Specifically, they often misinterpret branching instructions, omit critical IR operations, and rely on heuristic-based reasoning, leading to errors in CFG reconstruction, IR decompilation, and execution reasoning. The study underscores the necessity for IR-specific enhancements in LLMs, recommending fine-tuning on structured IR datasets and integration of explicit control flow models to augment their comprehension and handling of IR-related tasks.

Jiang, Hailong↗
DOE OSTI2025

AI for Interpreting Nuclear Power Plant Documents for Power Uprates

To reduce the cost and time needed for regulatory compliance, nuclear power plants (NPPs) can utilize artificial intelligence (AI) to assist in interpreting complex and voluminous documents that typically span thousands of pages. Usually, the process of interpreting a plant’s technical specifications (TSs) and associated documents is labor intensive. This study aims to understand what processes state-of-the-art large language models (LLMs) can automate and to identify the pitfalls associated with using LLMs to reduce human labor costs and time. This research uses a recent AI technology called retrieval augmented generation (RAG), which retrieves pages of information from TSs and associated documents to assist with NPP power uprates (cleared to produce more power). LLMs are integral to RAG because they create human-like responses based on the retrieved information, aiding in the interpretation and application processes. A baseline case demonstrates how LLMs can operate successfully for a power uprate application. Then five use cases show five types of potential failures: (1) RAG retrieving the incorrect information, (2) RAG misinterpreting the retrieved information, (3) RAG relying on knowledge not contained in the retrieved information, (4) RAG hallucinating, and (5) RAG refusing to answer. The results of the five use cases suggest that automating the human interpretation of TSs and associated documents with AI should be approached with caution. A subject-matter expert reviewed the AI outputs from the five use cases and concluded that an LLM can produce technical information that is needed to produce power uprate applications in certain instances.

21 - SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLAN↗
DOE OSTI2025

Leveraging large language models to automate the identification of healthcare access barriers for veterans

Objective: To develop and evaluate an automated system for identifying healthcare barriers focusing on transportation issues in veterans’ clinical notes using large language models (LLMs) and to assess the impact of different prompting strategies on classification performance and explanation consistency. Methods: We developed a hybrid system combining pattern matching for templated notes with LLM analysis for free-text notes. Using 2000 manually annotated clinical notes, we compared four prompting strategies (dual-role short, dual-role long, analysis-first, analysis-only) across Mistral-7B and Llama-3.1 models. We evaluated classification performance using standard metrics and assessed explanation consistency through embedding similarity analysis. Results: The analysis-first strategy achieved superior performance, with Mistral-7B reaching an F1 score of 0.914, outperforming traditional machine learning approaches (GBM: 0.786, BERT: 0.811). LLMs demonstrated higher explanation consistency within models (mean cosine similarity 0.887–0.908) compared to cross-model similarities (0.767–0.872). Pattern matching successfully handled 6.7% of templated notes deterministically. Mistral-7B showed greater internal consistency but higher abstention rates compared to Llama-3.1. Conclusion: Requiring LLMs to analyze evidence before classification improves both accuracy and explanation consistency for identifying transportation barriers in clinical notes. This approach enables automated barrier detection at scale while providing clinically relevant explanations, supporting both population-level healthcare planning and individual patient care decisions.

Healthcare access barriers↗
DOE OSTI2026

Leveraging Large Language Models for Understanding Fundamental Principles of Catalysis

Heterogeneous catalysis presents a distinct challenge for artificial intelligence (AI). Data sets are often small and inconsistently reported, catalyst representations are not standardized, and extracting fundamental knowledge requires integrating performance data, spectroscopic characterizations, and mechanistic models across multiple scales. Language offers a unifying representation across these modalities, making catalysis well suited for leveraging large language models (LLMs). By standardizing how catalytic data is represented, LLMs make dispersed experimental results more accessible to downstream statistical modeling. In this perspective, we focus our discussion around three opportunities where LLMs can significantly contribute to catalysis: (1) text to properties; (2) text to structure; and (3) text to mechanistic models. The discussion is followed by a perspective section on LLM-readiness of data, aligning LLM outputs with scientific correctness, and bridging lab-scale discovery to industrial deployment. Across each area, the most productive applications couple dispersed chemical knowledge with physics-grounded validation to produce verifiable hypotheses and actionable representations.

Catalysts↗
DOE OSTI2026

ChemGraph as an agentic framework for computational chemistry workflows

Atomistic simulations are essential in chemistry and materials science but remain challenging to run due to the expert knowledge required for the setup, execution, and validation stages of these calculations. We present ChemGraph, an agentic framework powered by artificial intelligence and state-of-the-art simulation tools to streamline and automate computational chemistry and materials science workflows. ChemGraph leverages graph neural network-based foundation models for accurate yet computationally efficient calculations and large language models (LLMs) for natural language understanding, task planning, and scientific reasoning to provide an intuitive and interactive interface. We evaluate ChemGraph across 13 benchmark tasks and demonstrate that smaller LLMs (GPT-4o-mini, Claude-3.5-haiku, Qwen-2.5-14B) perform well on simple workflows, while more complex tasks benefit from using larger models. Importantly, we show that decomposing complex tasks into smaller subtasks through a multi-agent framework enables GPT-4o to reach perfect accuracy and smaller LLMs to match or exceed single-agent GPT-4o's performance in these benchmarks.

Computational chemistry↗
DOE OSTI2026

A multimodal large language model for materials science

Understanding and predicting the properties of inorganic materials is crucial for accelerating advancements in materials science and driving applications in energy, electronics and beyond. Integrating material structure data with language-based information through multimodal large language models (LLMs) offers great potential to support these efforts by enhancing human–artificial intelligence interaction. However, a key challenge lies in integrating atomic structures at full resolution into LLMs. In this work, we introduce MatterChat, a versatile structure-aware multimodal LLM that unifies material structural data and textual inputs into a single cohesive model. MatterChat uses a bridging module to effectively align a pretrained universal machine learning interatomic potential with a pretrained LLM, reducing training costs and enhancing flexibility. Our results demonstrate that MatterChat greatly improves performance in material property prediction and human–artificial intelligence interaction, surpassing general-purpose LLMs such as GPT-4. We also demonstrate its usefulness in applications such as more advanced scientific reasoning and step-by-step material synthesis.

Tang, Yingheng [Lawrence Berkeley National Laborat↗
DOE OSTI2024

GraphAide: Advanced Graph-Assisted Query and Reasoning System

Curating knowledge from multiple siloed sources that contain both structured and unstructured data is a major challenge in many real-world applications. Pattern matching and querying represent fundamental tasks in modern data analytics that leverage this curated knowledge. The development of such applications necessitates overcoming several research challenges, including data extraction, named entity recognition, data modeling, and designing query interfaces. Moreover, the explainability of these functionalities is critical for their broader adoption. The emergence of Large Language Models (LLMs) has accelerated the development lifecycle of new capabilities. Nonetheless, there is an ongoing need for domain-specific tools tailored to user activities. The creation of digital assistants has gained considerable traction in recent years, with LLMs offering a promising avenue to develop such assistants utilizing domain-specific knowledge and assumptions. In this context, we introduce an advanced query and reasoning system, GraphAide, which constructs a knowledge graph (KG) from diverse sources and allows to query and reason over the resulting KG. GraphAide harnesses both the KG and LLMs to rapidly develop domain-specific digital assistants. It integrates design patterns from retrieval augmented generation (RAG) and the semantic web to create an agentic LLM application. GraphAide underscores the potential for streamlined and efficient development of specialized digital assistants, thereby enhancing their applicability across various domains.

Purohit, Sumit [BATTELLE (PACIFIC NW LAB)] (ORCID:↗
DOE OSTI2025

LLM-Inference-Bench: Inference Benchmarking of Large Language Models on AI Accelerators

Large Language Models (LLMs) have propelled groundbreaking advancements across several domains and are commonly used for text generation applications. However, the computational demands of these complex models pose significant challenges, requiring efficient hardware acceleration. Benchmarking the performance of LLMs across diverse hardware platforms is crucial to understanding their scalability and throughput characteristics. We introduce LLM-Inference-Bench, a comprehensive benchmarking suite to evaluate the hardware inference performance of LLMs. We thoroughly analyze diverse hardware platforms, including GPUs from Nvidia and AMD and specialized AI accelerators, Intel Habana and SambaNova. Our evaluation includes several LLM inference frameworks and models from LLaMA, Mistral, and Qwen families with 7B and 70B parameters. Our benchmarking results reveal the strengths and limitations of various models, hardware platforms, and inference frameworks. We provide an interactive dashboard to help identify configurations for optimal performance for a given hardware platform.

Chitty-Venkata, Krishna Teja↗
DOE OSTI2023

FORGE: Pre-Training Open Foundation Models for Science

Large language models (LLMs) are poised to revolutionize the way we conduct scientific research. However, both model complexity and pre-training cost are impeding effective adoption for the wider science community. Identifying suitable scientific use cases, finding the optimal balance between model and data sizes, and scaling up model training are among the most pressing issues that need to be addressed. In this study, we provide practical solutions for building and using LLM-based foundation models targeting scientific research use cases. We present an end-to-end examination of the effectiveness of LLMs in scientific research, including their scaling behavior and computational requirements on Frontier, the first Exascale supercomputer. We have also developed for release to the scientific community a suite of open foundation models called FORGE with up to 26B parameters using 257B tokens from over 200M scientific articles, with performance either on par or superior to other state-of-the-art comparable models. We have demonstrated the use and effectiveness of FORGE on scientific downstream tasks. Our research establishes best practices that can be applied across various fields to take advantage of LLMs for scientific discovery.

Yin, Junqi↗
DOE OSTI2025

Using a Large Language Model as a Building Block to Generate Usable Validation and Verification Suite for OpenMP

In the HPC area, both hardware and software move quickly. Often new hardware is developed and deployed, the corresponding software stack, including compilers and other tools, are under active development while leading edge software developers are working to port and tune their applications, all at the same time. While the software ecosystem is in flux, one of the key challenges for users is obtaining insight into the state of implementation of key features in the programming languages and models their applications are using – whether they have been implemented, and whether the implementation conforms to the specification, especially for newly implemented features (less tested by widespread use). OpenMP is one of the most prominent shared memory programming models used for on-node programming in HPC. With the shift towards accelerators (such as GPUs and FPGAs) and heterogeneous programming OpenMP features are getting more complex. It is natural to ask whether generative AI approaches, and large language models (LLMs) in particular, can help in producing validation and verification test suites to allow users better and faster insights into the availability and correctness of OpenMP features of interest. In this work, we explore the use of ChatGPT-4 to generate a suite of tests for OpenMP features. We have chosen a set of directives and clauses, a total of 78 combinations, which first appeared in OpenMP 3.0 (released in May 2008) but are also relevant for accelerators. We prompted ChatGPT to generate tests in the C and Fortran languages, for both host (CPU) and device (accelerator). On the Summit super-computer using the GNU implementation, we found that, of the 78 generated tests 67 C tests and 43 Fortran tests compiled successfully and fewer than those executed to completion. On further analysis we show that not all generated tests are valid. We document the process, results, and provide detailed analysis regarding the quality of tests generated. With the aim of providing input to a production quality validation and verification suite, we manually implement the corrections required to make the tests valid according to the current OpenMP specification. We quantify this effort as small, medium, or large, and record the lines of code changed to correct the invalid tests. With the corrected tests we validate recent implementations from HPE, AMD, and GNU on the Frontier supercomputer. Our experiment and subsequent analysis show that although LLMs are capable of producing HPC specific codes, they are limited by their understanding of the deeper semantics and restrictions of programming models such as OpenMP. Unsurprisingly more commonly used features have better support, while some OpenMP 3.0 directives such as sections and tasking are not universally supported on accelerators. We demonstrate that successful compilation and execution to completion are inadequate metrics for evaluating generated code and that, at this time, commodity LLMs require expert intervention for code verification. This points to gaps in the training data that is currently available for HPC. We demonstrate that with "small" effort 37% of generated invalid C tests and 63% of generated invalid Fortran tests could be corrected. This improves productivity of test generation as we circumvent writing from scratch and the common programming errors associated with it.

Pophale, Swaroop [ORNL] (ORCID:0000000185446367)↗
DOE OSTI2025

CAFE AU LAIT: Compute-Aware Federated Augmented Low-Rank AI Training

Federated finetuning is crucial for unlocking the knowledge embedded in pretrained Large Language Models (LLMs) when data are geographically distributed across clients. Unlike finetuning with data from a single institution, federated finetuning allows collaboration across multiple institutions, enabling the utilization of diverse and decentralized datasets while preserving data privacy. Given the high computing costs of LLM training and the emphasis on energy efficiency in Federated Learning (FL), Low-Rank Adaptation (LoRA) has emerged as a widely adopted algorithm due to its significantly reduced number of trainable parameters. However, this assumes that all data silos have the necessary computing resources to compute local updates of LLMs. Nevertheless, in practice, the computing resources across clients are highly heterogeneous: while some may have access to hundreds of GPUs, others might have limited or no GPU access. Recently, federated finetuning using synthetic data has been proposed, allowing clients to participate in a collaborative training run without training LLMs locally. However, our experimental results reveal a performance gap between models trained using synthetic data and those trained using local updates. Motivated by the observed heterogeneity in computing resources and the performance gap, we propose a novel two-stage algorithm that leverages the storage and computing capabilities of a strong server. In the first stage, under the coordination of the strong server, clients with limited computing resources collaborate to generate synthetic data, which is transferred to and stored on the strong server. In the second stage, the strong server uses this synthetic data on behalf of the resource-constrained clients to perform federated LoRA finetuning alongside clients with sufficient computing resources. This approach ensures that all clients can participate in the finetuning process. Experimental results demonstrate that incorporating local updates from even a small fraction of clients improves performance compared to using synthetic data for all clients. Furthermore, we incorporate the Gaussian mechanism in both stages to guarantee client-level differential privacy.

Wang, Jiayi [ORNL]↗
DOE OSTI2024

From Text to Maps: LLM-Driven Extraction and Geotagging of Epidemiological Data

Epidemiological datasets are essential for public health analysis and decision-making, yet they remain scarce and often difficult to compile due to inconsistent data formats, language barriers, and evolving political boundaries. Traditional methods of creating such datasets involve extensive manual effort and are prone to errors in accurate location extraction. To address these challenges, we propose utilizing large language models (LLMs) to automate the extraction and geotagging of epidemiological data from textual documents. Our approach significantly reduces the manual effort required, limiting human intervention to validating a subset of records against text snippets and verifying the geotagging reasoning, as opposed to reviewing multiple entire documents manually to extract, clean, and geotag. Additionally, the LLMs identify information often overlooked by human annotators, further enhancing the dataset’s completeness. Our findings demonstrate that LLMs can be effectively used to semi-automate the extraction and geotagging of epidemiological data, offering several key advantages: (1) comprehensive information extraction with minimal risk of missing critical details; (2) minimal human intervention; (3) higher-resolution data with more precise geotagging; and (4) significantly reduced resource demands compared to traditional methods.

Harrod, Karly↗
DOE OSTI2025

Connecting Minds: AI Use Cases to Bridge Power Systems and Large Language Models for Practical Applications

Recent advances in artificial intelligence (AI) and development of large language models (LLMs) present the opportunity to develop a new generation of power systems applications. In contrast with early power system AI applications based on structured numerical data, LLMs offer unique capabilities to perform logical reasoning using text documents, unstructured data, and application programming interface (API) calls to computational software. This paper seeks to bridge the knowledge gap between power systems engineers and LLM developers through a crosscutting explanation of use cases, characteristics, requirements, practical considerations from the perspectives of both LLM capabilities and industry needs. Specific focus is given to applications that can be realistically deployed by electric utilities. After introducing the architecture of LLMs and unique challenges of the power systems domain, this paper proposes twenty representative LLM applications grouped into categories of 1) power system operations, 2) asset management, 3) system planning and analytics, and 4) energy management and protection systems. Five use cases are presented within each category with descriptions of the motivation, objectives, approaches, example inputs / outputs, and benefits of each use case.

24 POWER TRANSMISSION AND DISTRIBUTION↗
DOE OSTI2024

A Framework for Evaluating the Implementation Cost of Attacks on Large Language Models

Large Language Models (LLMs) have been increasingly proposed as a method to enhance productivity in tasks that involve language and code. However, these models are large, complex, and their capabilities are not easily understood and controlled, meaning that their adoption opens many possibilities for new cyberattacks and misuse. Numerous attacks on LLMs have been reported and summarized in literature reviews, but we found existing reviews lacking in understanding the implementation cost of the attacks - i.e., how much effort would an attacker need in terms of coding, expertise, and resources to adopt attacks presented in the literature. Therefore, we divide existing attacks on LLMs into a taxonomy, and define a cost evaluation framework to determine the cost of the attack. An attack’s cost can be 1) estimated from reading the publication about the attack or 2) determined by implementing the attack from that publication. We provide an example evaluation of a couple jailbreaking frameworks based on experiments, and then apply the more lightweight cost estimate to a representative selection of attacks across the taxonomy we define. We discuss the relative difficulty of the attacks and also highlight defenses that have attempted to mitigate these attacks and assess their effectiveness.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Emergent Behaviors and the Limits of Large Language Model Generalization

Over the past couple years, large language models (LLMs) have rapidly grown in size and performance across wide numbers of tasks, leading some to herald them as machine learning models with truly “general” capabilities. We address some of the ways that the language used to describe these models can be both deceptive and a poor framing for making progress towards understanding their capabilities, and also review arguments on to what extent these models can represent and interact with the meaning of language. Emergent behaviors in these highly complex models are unpredictable and research into understanding them is still in early phases, which limits the strength of claims of generality about them. Moreover, while these models demonstrate surprising emergent reasoning capabilities on many tasks, there are still many limitations in generalization that overall high benchmark performance can hide. We review the literature showcasing limitations of LLMs as well as techniques to mitigate or overcome these challenges, while also highlighting how fundamental problems in model evaluation may prevent true claims of generality. We conclude with a practical section on using LLMs on real world problems while appropriately evaluating and validating these tasks to prevent or mitigate the downstream impacts of incorrect results.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Review of Grey Box/Black Box Data Contamination Metrics on Open and Commercial Models

Dataset contamination is a problem where benchmarks and tasks used to evaluate the capabilities of Large Language Models (LLMs) have been incorporated into the training dataset of the models. This gives a false sense of performance that can overestimate how these models will function on truly unseen data. This problem becomes worse with commercial LLMs with larger and non-accessible training data, so techniques have been developed to try to measure the degree to which a model is contaminated with a benchmark’s data. To understand the effectiveness of these techniques, particularly when evaluating contamination on coding tasks, we review trends and categorize techniques by the degree of access to the model that is required. The research literature on this topic has reported mixed effectiveness of these techniques, so we select a set of black box (text access only) and grey box (access to model loss/probabilities required) techniques and apply them to both commercial and non-commercial models. We implement these metrics as part of a framework to test the contamination of Python code in LLMs to see to what extent we can replicate the effectiveness (or ineffectiveness) of these contamination detection techniques. Though we find mixed results in the capabilities of these metrics to identify contamination, we do observe evidence that they can identify contamination (broadly) in fine-tuned models when both a baseline and fine-tuned model is present. Additionally, similarity metrics were able to identify between contaminated and uncontaminated data even in situations where the data is distributionally similar (e.g., drawn from the same set of code projects).

97 MATHEMATICS AND COMPUTING↗