SEARCH · Search NASA

Results for “Small Language Model”

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.

SourceResearch connectionFrom yearThrough year

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

At least 109 records · Page 6

All sources NASA NTRS DOE OSTI

DOE OSTI2025

AUTOMATIC GENERATION OF EVENT TREES AND FAULT TREES: A MODEL-BASED APPROACH

In the past few decades, increasing complexity in modern engineering systems has been driven by the integration of a large number of components and by the fact that the system operations involve many disciplines (e.g., thermal-hydraulics, plant operations, cyber-security). Current safety/reliability modeling approaches to such systems are labor intensive, difficult to learn, and rely heavily on simplistic Boolean logic to depict failure propagation and accident progression. While these methods serve well for simple systems (i.e., linear causal systems with limited small inter- and intra-system interactions), their results are difficult to verify when modeling complex systems (typically performed through the extensive use of modeling assumptions). The development of new methods is addressed to meet these challenges through a model-based system engineering (MBSE) lens. Under MBSE philosophy, every aspect of the system (form or function) is represented by a model that completely characterizes its architecture or behavior. MBSE approach greatly improves the management of design, analysis and verification of complex systems. An integration of Dynamic Probabilistic Risk Assessment (DPRA) methods with MBSE models is proposed to perform safety/reliability analyses of engineering systems. In particular, MBSE representation of the system (performed using Systems Modeling Language [SysML]) is coupled with DPRA methods to automatically generate event trees and fault trees.

97 - MATHEMATICS AND COMPUTING↗

DOE OSTI2026

Foundation models for atomistic simulation of chemistry and materials

Conventional computational methods for modeling chemical and materials systems are limited by system size and timescale, forcing a trade-off between quantum-mechanical accuracy and the sampling needed for realistic observables. Large language and vision foundation models — pre-trained on massive datasets using transformer architectures — have revolutionized many fields. It is thus interesting to ask whether a foundation model — subject to suitable data, parameter scaling and training — could enable learned simulations of chemistry and materials. Here, in this study, we review the field of machine-learned interatomic potentials (MLIPs) and posit that scaling up large and diverse chemical and materials datasets and highly expressive architectures using advanced training strategies should result in models that are: more efficient, transferable, robust to out-of-distribution scenarios, and easier to fine-tune to a variety of downstream physical observables than models trained from scratch on small datasets corresponding to specific, targeted atomistic simulation tasks. We provide specific criteria for creating such large-scale MLIP foundation models, coordinated strategies for their development, evaluation and deployment, and highlight potential emergent capabilities that could transform predictive simulations in chemistry and materials science and accelerate discovery across multiple technological domains.

Yuan, Eric C.-Y. [University of California, Berkel↗

NASA NTRS2018

Using Board Games as Subject Matter for Developing Expertise in Model-Based Systems Engineering

As more organizations transition from traditional document-centric systems engineering to a model-based approach, many are challenged to train their staff in new languages, tools, and methodologies, while managing the expectations of stakeholders and their expected model outcomes. In particular, challenges associated with learning a new modeling language and developing skills in the 'art' of modeling present organizations with formidable obstacles to realizing this transition. This paper hypothesizes that systems engineers may more readily learn how to correctly model with SysML, and develop intuition about the art of modeling and using patterns, if their learning references a commonly and thoroughly-understood subject, such as a board game. This paper presents a case for the use of board games as subject matter for new modelers. It demonstrates the concept with a sample model of Hasbro's popular board game, Monopoly, and discusses the limitations of this approach and potential adaptations that may broaden the applicability of the learned skills to projects. Finally, results from a small feasibility assessment and concepts for more formal study to evaluate the hypothesis are presented.

Model-Based Systems Engineering↗

NASA NTRS2002

REACH: Real-Time Data Awareness in Multi-Spacecraft Missions

Missions have been proposed that will use multiple spacecraft to perform scientific or commercial tasks. Indeed, in the commercial world, some spacecraft constellations already exist. Aside from the technical challenges of constructing and flying these missions, there is also the financial challenge presented by the tradition model of the flight operations team (FOT) when it is applied to a constellation mission. Proposed constellation missions range in size from three spacecraft to more than 50. If the current ratio of three-to-five FOT personnel per spacecraft is maintained, the size of the FOT becomes cost prohibitive. The Advanced Architectures and Automation Branch at the Goddard Space Flight Center (GSFC Code 588) saw the potential to reduce the cost of these missions by creating new user interfaces to the ground system health-and-safety data. The goal is to enable a smaller FOT to remain aware and responsive to the increased amount of ground system information in a multi-spacecraft environment. Rather than abandon the tried and true, these interfaces were developed to run alongside existing ground system software to provide additional support to the FOT. These new user interfaces have been combined in a tool called REACH. REACH-the Real-time Evaluation and Analysis of Consolidated Health-is a software product that uses advanced visualization techniques to make spacecraft anomalies easy to spot, no matter how many spacecraft are in the constellation. REACH reads a real-time stream of data from the ground system and displays it to the FOT such that anomalies are easy to pick out and investigate. Data visualization has been used in ground system operations for many years. To provide a unique visualization tool, we developed a unique source of data to visualize: the REACH Health Model Engine. The Health Model Engine is rule-based software that receives real-time telemetry information and outputs "health" information related to the subsystems and spacecraft that the telemetry belong to. The Health Engine can run out-of-the-box or can be tailored with a scripting language. Out of the box, it uses limit violations to determine the health of subsystems and spacecraft; when tailored, it determines health using equations combining the values and limits of any telemetry in the spacecraft. The REACH visualizations then "roll up" the information from the Health Engine into high level, summary displays. These summary visualizations can be "zoomed" into for increasing levels of detail. Currently REACH is installed in the Small Explorer (SMEX) lab at GSFC, and is monitoring three of their five spacecraft. We are scheduled to install REACH in the Mid-sized Explorer (MIDEX) lab, which will allow us to monitor up to six more spacecraft. The process of installing and using our "research" software in an operational environment has provided many insights into which parts of REACH are a step forward and which of our ideas are missteps. Our paper explores both the new concepts in spacecraft health-and-safety visualization, the difficulties of such systems in the operational environment, and the cost and safety issues of multi-spacecraft missions.

Maks, Lori↗

DOE OSTI2025

ChatHPC: Building the Foundations for a Productive and Trustworthy AI-Assisted HPC Ecosystem

ChatHPC democratizes large language models for the high-performance computing (HPC) community by providing the infrastructure, ecosystem, and knowledge needed to apply modern generative AI technologies to rapidly create specific capabilities for critical HPC components while using relatively modest computational resources. Our divide-and-conquer approach focuses on creating a collection of reliable, highly specialized, and optimized AI assistants for HPC based on the cost-effective and fast Code Llama fine-tuning processes and expert supervision. We target major components of the HPC software stack, including programming models, runtimes, I/O, tooling, and math libraries. Thanks to AI, ChatHPC provides a more productive HPC ecosystem by boosting important tasks related to portability, parallelization, optimization, scalability, and instrumentation, among others. With relatively small datasets (on the order of KB), the AI assistants, which are created in a few minutes by using one node with two NVIDIA H100 GPUs and the ChatHPC library, can create new capabilities with Meta’s 7-billion parameter Code Llama base model to produce high-quality software with a level of trustworthiness of up to 90% higher than the 1.8-trillion parameter OpenAI ChatGPT-4o model for critical programming tasks in the HPC software stack.

Young, Aaron [ORNL] (ORCID:0000000254484667)↗

NASA NTRS2023

A Common Trajectory Language for New Airspace Domains

Revolutionary new aviation services are currently in development, including small package delivery and short-range urban passenger transportation. These new services need a language to represent and deconflict trajectories, and a common language will simplify coordination between the new airspace domains as well as traditional airspace. A Trajectory Specification Language (TSL) was previously proposed for traditional air traffic, and it can also serve as a common language for the new aviation domains and services. The TSL specifies a reference trajectory in 4D space along with spatial tolerances that determine a bounding volume at any given time in flight. The tolerances are defined in terms of the route-oriented cross-track, along-track, and vertical axes, and they are allowed to vary with distance along the route. This bounding model can guarantee safe separation between flights as long as they are in conformance with their assigned trajectories. The advantages of this model over the previously proposed Operational Intent Volumes (OIVs) in terms of airspace usage efficiency are explained.

air traffic control, aircraft trajectory, language↗

NASA NTRSDate not supplied

Document Classification Techniques for Aviation Letters of Agreement

Often when working with historic air traffic management (ATM) documents, it is helpful to classify them into specific categories. In this paper, we conduct a thorough review of natural language processing techniques to perform this classification task on Letters of Agreement (LOAs), technical aviation documents outlining rules for utilizing US airspace. We evaluate multiple techniques for representing the text in the documents as embeddings: unigram and bigram Term Frequency Inverse Document Frequency (TFIDF), Word2Vec, Doc2Vec, GloVe and RoBERTa. We investigate a wide range of classification models: K-Nearest Neighbors, Random Forest, Support Vector Machines (SVM), Logistic Regression, Naive Bayes, Feed-Forward Neural Network, Convolutional Neural Networks (CNNs) and Long-Short Term Memory (LSTM). By comparing the different methods, we found the best overall approach for our task was to use unigram TFIDF representations with SVM while also gaining insight into how the other methodologies performed on a small technical datasets.

ATM↗

NASA NTRSDate not supplied

Document Classification Techniques for Aviation Letters of Agreement

ATM↗

NASA NTRSDate not supplied

Multi-Domain Routing in Delay Tolerant Networks

The goal of Delay Tolerant Networking (DTN) is to provide the missing ingredient for the ever-growing collection of communicating nodes in our solar system to become a Solar System Internet (SSI). Great strides have been made in modeling particular types of DTNs, such as schedule- or discovery-based. Now, analogously to the Internet, these smaller DTNs can be considered routing domains which must be stitched together to form the overall SSI. In this paper, we propose a framework for cross-domain routing in DTNs as well as methodologies for detecting these sub-domains. Example time-varying networks are given to demonstrate the techniques proposed. A basic component is the mathematical theory of sheaves, which unifies the underlying model of DTN routing algorithms, by giving rise to routing sheaves – these can be defined for the dynamic and scheduled networks as noted above, and can also be used to define the interfaces between these domains in order to route across them. An immediate application would be routing across discovery-based networks connected by scheduled networks. These DTN subdomains remain elusive, however, and need to become well-defined and properly sized for tractable computability. In particular, a balance must be determined between areas that are too large (i.e. large matrix computations) versus areas that are too small (i.e. “many” single-noded domains). Moreover, the connections between the domains should, at least locally, be chosen to optimize data flow and connectivity: we address this in three ways. First, tools from persistent homology are given to understand underlying structures, reminiscent of hierarchies in the Internet Protocol (IP) addressing. Second, we construct a notion of temporal graph curvature based on network geometry to analyze flows induced by dynamical processes on these networks. Finally, Schrodinger Bridges, a tool arising from statistical physics, are proposed as a method of constructing flows on time-evolving networks with desirable properties such as speed, robustness, and load sensitivity. We construct an approach to temporal hypergraphs to simultaneously model unicast, multicast, and broadcast, using the language of scheme theory, and then consider DTN network coding as a way to achieve network-level computation and organization. The paper concludes with a discussion and ideas for future work.

Alan Hylton↗

NASA NTRSDate not supplied

Document Classification Techniques for Aviation Letters of Agreement

Often when working with technical documents, it is helpful to classify them into specific categories. In this paper, we conduct a thorough review of natural language processing techniques to perform this classification task on Letters of Agreement (LOAs), technical aviation documents outlining rules for utilizing US airspace. We evaluate multiple techniques, including Transfer Learning, for representing the text in the documents as embeddings: unigram and bigram Term Frequency Inverse Document Frequency (TFIDF), Word2Vec, Doc2Vec, GloVe and RoBERTa. We investigate a wide range of classification models: K-Nearest Neighbors, Random Forest, Support Vector Machines (SVM), Logistic Regression, Naive Bayes, Feed-Forward Neural Network, Convolutional Neural Networks (CNNs) and Long-Short Term Memory (LSTM). By comparing the different methods, we found the best overall approach for our task was to use unigram TFIDF representations with SVM while also gaining insight into how the other methodologies performed on a small technical datasets.

Aayushi Batra↗

DOE OSTI2022

Multinode Multi-GPU Two-Electron Integrals: Code Generation Using the Regent Language

The computation of two-electron repulsion integrals (ERIs) is often the most expensive step of integral-direct self-consistent field methods. Formally it scales as O(N 4 ), where N is the number of Gaussian basis functions used to represent the molecular wave function. In practice, this scaling can be reduced to O(N 2 ) or less by neglecting small integrals with screening methods. The contributions of the ERIs to the Fock matrix are of Coulomb (J) and exchange (K) type and require separate algorithms to compute matrix elements efficiently. We previously implemented highly efficient GPU-accelerated J-matrix and K-matrix algorithms in the electronic structure code TeraChem. Although these implementations supported the use of multiple GPUs on a node, they did not support the use of multiple nodes. This presents a key bottleneck to cutting-edge ab initio simulations of large systems, e.g., excited state dynamics of photoactive proteins. We present our implementation of multinode multi-GPU J- and K-matrix algorithms in TeraChem using the Regent programming language. Regent directly supports distributed computation in a task-based model and can generate code for a variety of architectures, including NVIDIA GPUs. We demonstrate multinode scaling up to 45 GPUs (3 nodes) and benchmark against hand-coded TeraChem integral code. Finally, we also outline our metaprogrammed Regent implementation, which enables flexible code generation for integrals of different angular momenta.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

NASA NTRS2010

Autonomous GN and C for Spacecraft Exploration of Comets and Asteroids

A spacecraft guidance, navigation, and control (GN&C) system is needed to enable a spacecraft to descend to a surface, take a sample using a touch-and-go (TAG) sampling approach, and then safely ascend. At the time of this reporting, a flyable GN&C system that can accomplish these goals is beyond state of the art. This article describes AutoGNC, which is a GN&C system capable of addressing these goals, which has recently been developed and demonstrated to a maturity TRL-5-plus. The AutoGNC solution matures and integrates two previously existing JPL capabilities into a single unified GN&C system. The two capabilities are AutoNAV and GREX. AutoNAV is JPL s current flight navigation system, and is fairly mature with respect to flybys and rendezvous with small bodies, but is lacking capability for close surface proximity operations, sampling, and contact. G-REX is a suite of low-TRL algorithms and capabilities that enables spacecraft operations in close surface proximity and for performing sampling/contact. The development and integration of AutoNAV and G-REX components into AutoGNC provides a single, unified GN&C capability for addressing the autonomy, close-proximity, and sampling/contact aspects of small-body sample return missions. AutoGNC is an integrated capability comprising elements that were developed separately. The main algorithms and component capabilities that have been matured and integrated are autonomy for near-surface operations, terrain-relative navigation (TRN), real-time image-based feedback guidance and control, and six degrees of freedom (6DOF) control of the TAG sampling event. Autonomy is achieved based on an AutoGNC Executive written in Virtual Machine Language (VML) incorporating high-level control, data management, and fault protection. In descending to the surface, the AutoGNC system uses camera images to determine its position and velocity relative to the terrain. This capability for TRN leverages native capabilities of the original AutoNAV system, but required advancements that integrate the separate capabilities for shape modeling, state estimation, image rendering, defining a database of onboard maps, and performing real-time landmark recognition against the stored maps. The ability to use images to guide the spacecraft requires the capability for image-based feedback control. In Auto- GNC, navigation estimates are fed into an onboard guidance and control system that keeps the spacecraft guided along a desired path, as it descends towards its targeted landing or sampling site. Once near the site, AutoGNC achieves a prescribed guidance condition for TAG sampling (position/orientation, velocity), and a prescribed force profile on the sampling end-effector. A dedicated 6DOF TAG control then implements the ascent burn while recovering from sampling disturbances and induced attitude rates. The control also minimizes structural interactions with flexible solar panels and disallows any part of the spacecraft from making contact with the ground (other than the intended end-effector).

Carson, John M.↗

NASA NTRS2010

AutoGNC Testbed

A simulation testbed architecture was developed and implemented for the integration, test, and development of a TRL-6 flight software set called Auto- GNC. The AutoGNC software will combine the TRL-9 Deep Impact AutoNAV flight software suite, the TRL-9 Virtual Machine Language (VML) executive, and the TRL-3 G-REX guidance, estimation, and control algorithms. The Auto- GNC testbed was architected to provide software interface connections among the AutoNAV and VML flight code written in C, the G-REX algorithms in MATLAB and C, stand-alone image rendering algorithms in C, and other Fortran algorithms, such as the OBIRON landmark tracking suite. The testbed architecture incorporates software components for propagating a high-fidelity truth model of the environment and the spacecraft dynamics, along with the flight software components for onboard guidance, navigation, and control (GN&C). The interface allows for the rapid integration and testing of new algorithms prior to development of the C code for implementation in flight software. This testbed is designed to test autonomous spacecraft proximity operations around small celestial bodies, moons, or other spacecraft. The software is baselined for upcoming comet and asteroid sample return missions. This architecture and testbed will provide a direct improvement upon the onboard flight software utilized for missions such as Deep Impact, Stardust, and Deep Space 1.

Carson, John M., III↗

DOE OSTI2025

Enhancing ChatPORT with CUDA-to-SYCL Kernel Translation Capability

Large Language Models (LLMs) have shown strong capabilities in general code translation. However, code translation involving parallel programming models remains largely unexplored. This work enhances the capabilities of code LLMs in CUDA-to-SYCL kernel translation with parameter-efficient fine-tuning. The resultant fine-tuned LLM, called ChatPORT, is an effort to provide high-fidelity translations from one programming model to another. We describe the preparation of datasets from heterogeneous computing benchmarks for model fine-tuning and testing, the parameter-efficient fine-tuning of 19 open-source code models ranging in size from 0.5 to 34 billion parameters and evaluate the correctness rates of the SYCL kernels by the fine-tuned models. The experimental results show that most code models fail to translate CUDA codes to SYCL correctly. However, fine-tuning these models using a small set of CUDA and SYCL kernels can enhance the capabilities of these models in kernel translation. Depending on the sizes of the models, the correctness rate ranges from 19.9% to 81.7% for a test dataset of 62 CUDA kernels.

Jin, Zheming [ORNL] (ORCID:000000027197780X)↗

NASA NTRS2001

A Scalable Analysis Toolkit

The Scalable Analysis Toolkit (SAT) project aimed to demonstrate that it is feasible and useful to statically detect software bugs in very large systems. The technical focus of the project was on a relatively new class of constraint-based techniques for analysis software, where the desired facts about programs (e.g., the presence of a particular bug) are phrased as constraint problems to be solved. At the beginning of this project, the most successful forms of formal software analysis were limited forms of automatic theorem proving (as exemplified by the analyses used in language type systems and optimizing compilers), semi-automatic theorem proving for full verification, and model checking. With a few notable exceptions these approaches had not been demonstrated to scale to software systems of even 50,000 lines of code. Realistic approaches to large-scale software analysis cannot hope to make every conceivable formal method scale. Thus, the SAT approach is to mix different methods in one application by using coarse and fast but still adequate methods at the largest scales, and reserving the use of more precise but also more expensive methods at smaller scales for critical aspects (that is, aspects critical to the analysis problem under consideration) of a software system. The principled method proposed for combining a heterogeneous collection of formal systems with different scalability characteristics is mixed constraints. This idea had been used previously in small-scale applications with encouraging results: using mostly coarse methods and narrowly targeted precise methods, useful information (meaning the discovery of bugs in real programs) was obtained with excellent scalability.

Aiken, Alexander↗

DOE OSTI2024

Assessment of fine-tuned large language models for real-world chemistry and material science applications

The current generation of large language models (LLMs) has limited chemical knowledge. Recently, it has been shown that these LLMs can learn and predict chemical properties through fine-tuning. Using natural language to train machine learning models opens doors to a wider chemical audience, as field-specific featurization techniques can be omitted. In this work, we explore the potential and limitations of this approach. We studied the performance of fine-tuning three open-source LLMs (GPT-J-6B, Llama-3.1-8B, and Mistral-7B) for a range of different chemical questions. We benchmark their performances against “traditional” machine learning models and find that, in most cases, the fine-tuning approach is superior for a simple classification problem. Depending on the size of the dataset and the type of questions, we also successfully address more sophisticated problems. The most important conclusions of this work are that, for all datasets considered, their conversion into an LLM fine-tuning training set is straightforward and that fine-tuning with even relatively small datasets leads to predictive models. These results suggest that the systematic use of LLMs to guide experiments and simulations will be a powerful technique in any research study, significantly reducing unnecessary experiments or computations.

Van Herck, Joren↗

DOE OSTI2021

Re-evaluation of ortho-para- dependence of self pressure-broadening in the ν 1 + ν 3 band of acetylene

Optical frequency comb-referenced measurements of self pressure-broadened line profiles of the R(8) to R(13) lines in the ν 1 + ν 3 combination band of acetylene near 1.52 µ m are reported. The analysis of the data found no evidence for a previously reported systematic alternation in self pressure-broadened line widths with the nuclear spin state of the molecule. This work brought out the need for the use of an accurate line profile model and careful accounting for weak background absorptions due to hot band and lower abundance isotopomer lines. The data were adequately fit using the quadratic speed-dependent Voigt profile model, neglecting the small speed-dependent shift. Parameters describing the most probable and speed-dependent pressure-broadening, most probable shift, and the line strength were determined for each line. Detailed modeling of the results of Iwakuni et al. showed that their neglect of collisional narrowing due to the speed-dependent broadening term combined with the strongly absorbing data recorded and analyzed in transmission mode were the reasons for their results.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗

NASA NTRSDate not supplied

Digitizing Named Entities Found Within Letters of Agreement

Letters of Agreement (LOAs) are text-based air traffic control documents that contain procedures and actions agreed upon by the different parties, typically two or more FAA facilities, that are subject to an agreement. The documents contain among other things generic constraints, which are explicit and implicit combinations of procedures that limit a flight’s trajectory and affects pilot actions. For example, a controller may be required, to assign a specific altitude to an aircraft crossing the boundary between two airspaces. Although LOA generic constraints directly impact the trajectory of an aircraft, they are not currently available in a digital form that can be used for (or directly ingested into automated) flight planning. Instead, the constraints are manually input into an onboard or ground based system. LOA documents are primarily stored at a controlling facility and the generic constraints are implemented by experienced air traffic controllers and pilots primarily using voice instructions. This increases the workload of the controllers, likelihood of error (e.g., due to noisy communication) and makes it impractical for implementation with unmanned aircraft. Therefore, steps must be taken to make existing constraints machine interpretable to enable e.g., automated handoffs which in turn would reduce controller workload. With recent advances in natural language processing, especially the rise in digitization of text documents (e.g., medical documents) and automated extraction of information therein, it is now possible to extract flight specific constraints from LOAs. The goal of this work is to digitize named entities through a combination of natural language processing tasks: named entity disambiguation, toponym resolution, and numeric parsing to extract general constraint components contained within LOAs, herein referred to as Entity Enhancement (EE). Starting with a small list of named entities (e.g., ARTCC, Tower, Altitude and Speed), EE can extract the named entities while simultaneously converting the string-based output into a digital format using an ensemble of processes like rule-based gazetteers and syntactic-lexical patterns. The digital format contains a diverse set of information based on the entity label in question, ranging from standardized facility names to units of measure (e.g., feet) and other numeric information. Upon validating our approach using a truth dataset, we show an overall F1-Score of 0.71 for the extraction process. Looking beyond entity enhancement, we are also working towards the goal of completely digitizing the general constraints by performing EE and fitting them into a standardized exchange model (XM) such as the Aeronautical Information Exchange Model (AIXM). This will allow for easy distribution and dissemination of LOA constraints to air users, better searchability within documents, and enable ingestion into automated flight planning. Finally, we show a preliminary version of the proposed XM architecture and demonstrate how the model can be populated from the EE output.

Stephen S. B. Clarke↗