Search NASA⌕ Search

SEARCH · Search NASA

Results for “text corpus”

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.

36 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2024

Understanding Event Trajectories Across Massive Temporal Datasets with Word Embeddings and Visualization

In collaboration with researchers from Virginia Tech, Savannah River National Laboratory has continued development of a natural language processing pipeline to identify and extract events of interest from massive open data sources in the domain of worldwide state-sponsored civil nuclear energy. The foundation of the pipeline is built on compass aligned temporal word embedding models, whereby contextual shifts are automatically identified by comparing keyword embedding vectors across successive time windows. Within the approach, a contextual shift indicates the occurrence of a potential event of interest. However, in such a broad topical domain that captures events at a global scale, across various life cycle stages, and across numerous different technology types, a user that is monitoring events may have broad interests in capturing many different event types with varying degrees of signal. As such, the quantity of information that may be returned from an automated event extraction pipeline can be substantial, requiring manual effort to sift through the information to identify any relevant bits of information. Therefore, a more streamlined workflow that aids in directing a user toward specific information at different points in time is necessary. The workflow presented here has been developed with this concept in mind, built on top of the initial prototype event extraction pipeline, whereby a user can analyze temporal text-based data sources at multiple different contextual levels to isolate key points in time and key subdomains captured within a data corpus. Using multiple corpuses that consist of approximately 7 million Tweets and 7 million news articles, the team has extended compass aligned temporal word embedding models to establish an interconnected and hierarchical structure that relates known key words of interest to documents, local topics (i.e., within a time window), and global topics across the corpuses. All of this information is packaged into a visual analytics system that is linked to the information extraction pipeline and enables a user to identify contextual information that describes the evolution of a high dimensional embedding space across time to isolate changes of interest and explore associated events. This report demonstrates the use of these analytics and a means to fuse information across multiple datasets.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Hypercane

Automatically summarizing a collection of documents requires analyzing their features using various operations. Hypercane is a framework for building algorithms for sampling files from a collection. Sampling is a vital part of automatic summarization. Hypercane ties together many existing third-party libraries. These libraries include those for processing web resources, natural language processing for analyzing text and filtering documents, and machine learning for clustering summarization candidates. A user can run up to more than 70 operations on a corpus in the order of their choosing. Through these operations, they can create custom automatic summarization algorithms to help them process the corpus of their choice. Details of the algorithms possible with Hypercane are available in Jones 2021. Details of running Hypercane are available in Jones et al.; 2021.

Jones, Shawn↗
DOE OSTI2026

Natural Language Processing to Inform Agent-Based Modeling: With Application to Modeling Adoption of Medium-Duty Electric Vehicles

Agent-based socio-technical modeling of medium- and heavy-duty (MDHD) electric vehicle (EV) adoption has the potential to provide analysis, prediction, and gui. This paper describes new applications of text analysis developed through machine learning (ML) to build and understand relevant topics and their saliency in the published discourse on adoption of MDHD EVs. This work contributes to the state of the art in topic mining models by defining a new metric of topic ranking (START) that quantifies the importance of predefined topics within the corpus using weighted results for predefined topics from two topic modeling approaches: Latent Dirichlet Allocation (LDA) and BERTopic. The START metric is then demonstrated in practice to model how academia and industry view the EV adoption process based on the respective texts published by these groups. Results show that academic literature places more emphasis on categories of interests such as norms/attitudes and adopter knowledge, while trade journals tend to emphasize long-term cost more than academia. The two bodies of literature agree on the importance of policy and incentives in MDHD EV adoption. Together these results illustrate the potential to use ML-based text analysis to populate the characteristics of agent-based socio-technical models.

Electric vehicle adoption, fleet electrification, ↗
DOE OSTI2020

JSONize: A Scalable Machine Learning Pipeline to Model Medical Notes as Semi-structured Documents

The Department of Veteran's Affairs (VA) archives the largest corpora of clinical notes in their corporate data warehouse (CDW) as unstructured text data. Unstructured text easily supports keyword searches and regular expressions. Often these simple searches do not adequately support the complex searches that need to be performed on notes. For example, a researcher may want all notes with a Duke Treadmill Score less than 5 or people that smoke more than 1 pack per day. Range queries like this and more can be supported by modelling text as semi-structured documents. In this paper, we implement a scalable machine learning pipeline that models plain medical text as useful semi-structured documents. We improve on existing models and achieve a F1-score of 0.912 and scale our methods to the entire VA corpus.

Rush III, Everett↗
DOE OSTI2020

Hypergraph Random Walks, Laplacians, and Clustering

We propose a flexible framework for clustering hypergraph-structured data based on recently proposed random walks utilizing edge-dependent vertex weights. When incorporating edge-dependent vertex weights (EDVW), a weight is associated with each vertex-hyperedge pair, yielding a weighted incidence matrix of the hypergraph. Such weightings have been utilized in term-document representations of text data sets. We explain how random walks with EDVW serve to construct different hypergraph Laplacian matrices, and then develop a suite of clustering methods that use these incidence matrices and Laplacians for hypergraph clustering. Using 20Newsgroup, U.S. patent, Reuters' Corpus Volume 1, and genetics data sets, we compare the performance of these clustering algorithms experimentally against a variety of existing hypergraph clustering methods. We show that the proposed methods produce higher-quality clusters.

hypergraphs, clustering, laplacian, random walk, M↗
DOE OSTI2026

MOSAIC-CONUS: A Multimodal, Multi-Temporally Paired Dataset for Earth Sciences

Earth embeddings—vector representations of geographic locations indexed in space and time—are emerging as a unifying interface for geospatial AI. However, their quality depends not only on model design, but on how multimodal Earth observation (EO) data are spatially indexed, temporally aligned, and cross-modally associated during pretraining. We introduce MOSAIC-CONUS (Multimodal Observations with Spatially Aligned Imagery, Urban Points of Interest, In-Situ Measurements and Text Captions), a large-scale EO dataset over the contiguous United States, organized around 250,000 stratified point indices that serve as stable spatial keys across seven modalities: active radar, passive optical imagery, lidar-derived elevation, land cover, functional context, hydrometeorological measurements, and textual summaries. Unlike existing EO datasets, MOSAIC-CONUS introduces four contributions not jointly addressed in prior work: 1. an open-source, large-scale multimodal EO corpus structured around point-indexed data designed to support Earth embedding learning; 2. explicit radar-optical pairing tables spanning twelve temporal alignment regimes, formalizing cross-sensor alignment as a controllable variable for analyzing how temporal mismatch across modalities influences learned embeddings quality; 3. a benchmark suite spanning cross-modal retrieval, annual nightlights regression, and basin-held-out streamflow prediction, positioning MOSAIC-CONUS as a benchmark-ready resource for multimodal AI systems; and 4. a language-based embedding layer through co-registered textual summaries, enabling Earth embeddings to function as a queryable interface for agentic AI systems. The dataset and pairing protocols are publicly released.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2021

Models and Processes to Extract Drug-like Molecules From Natural Language Text

Researchers worldwide are seeking to repurpose existing drugs or discover new drugs to counter the disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). A promising source of candidates for such studies is molecules that have been reported in the scientific literature to be drug-like in the context of viral research. However, this literature is too large for human review and features unusual vocabularies for which existing named entity recognition (NER) models are ineffective. We report here on a project that leverages both human and artificial intelligence to detect references to such molecules in free text. We present 1) a iterative model-in-the-loop method that makes judicious use of scarce human expertise in generating training data for a NER model, and 2) the application and evaluation of this method to the problem of identifying drug-like molecules in the COVID-19 Open Research Dataset Challenge (CORD-19) corpus of 198,875 papers. We show that by repeatedly presenting human labelers only with samples for which an evolving NER model is uncertain, our human-machine hybrid pipeline requires only modest amounts of non-expert human labeling time (tens of hours to label 1778 samples) to generate an NER model with an F-1 score of 80.5%—on par with that of non-expert humans—and when applied to CORD’19, identifies 10,912 putative drug-like molecules. This enriched the computational screening team’s targets by 3,591 molecules, of which 18 ranked in the top 0.1% of all 6.6 million molecules screened for docking against the 3CLPro protein.

60 APPLIED LIFE SCIENCES↗
DOE OSTI2025

Consistent performance of large language models in rare disease diagnosis across ten languages and 4917 cases

Background Large language models (LLMs) are increasingly used medicine for diverse applications including differential diagnostic support. The training data used to create LLMs such as the Generative Pretrained Transformer (GPT) predominantly consist of English-language texts, but LLMs could be used across the globe to support diagnostics if language barriers could be overcome. Initial pilot studies on the utility of LLMs for differential diagnosis in languages other than English have shown promise, but a large-scale assessment on the relative performance of these models in a variety of European and non-European languages on a comprehensive corpus of challenging rare-disease cases is lacking. Methods We created 4917 clinical vignettes using structured data captured with Human Phenotype Ontology (HPO) terms with the Global Alliance for Genomics and Health (GA4GH) Phenopacket Schema. These clinical vignettes span a total of 360 distinct genetic diseases with 2525 associated phenotypic features. We used translations of the Human Phenotype Ontology together with language-specific templates to generate prompts in English, Chinese, Czech, Dutch, French, German, Italian, Japanese, Spanish, and Turkish. We applied GPT-4o, version gpt-4o-2024-08-06, and the medically fine-tuned Meditron3-70B to the task of delivering a ranked differential diagnosis using a zero-shot prompt. An ontology-based approach with the Mondo disease ontology was used to map synonyms and to map disease subtypes to clinical diagnoses in order to automate evaluation of LLM responses. Findings For English, GPT-4o placed the correct diagnosis at the first rank 19.9% and within the top-3 ranks 27.0% of the time. In comparison, for the nine non-English languages tested here the correct diagnosis was placed at rank 1 between 16.9% and 20.6%, within top-3 between 25.4% and 28.6% of cases. The Meditron3 model placed the correct diagnosis within the first 3 ranks for 20.9% of cases in English and between 19.9% and 24.0% for the other nine languages. Interpretation The differential diagnostic performance of LLMs across a comprehensive corpus of rare-disease cases was largely consistent across the ten languages tested. This suggests that the utility of LLMs in clinical settings may extend to non-English clinical settings.

Artificial intelligence↗
DOE OSTI2022

Machine Learning Modeling Pipeline for Extracting Nuclear Proliferation Events of Interest from Open Data Sources (U)

In FY2020, the Savannah River National Laboratory (SRNL) and the Sanghani Center for Artificial Intelligence and Data Analytics at Virginia Polytechnic Institute and State University entered a collaboration funded by Department of Energy’s (DOE) Office of Defense Nuclear Nonproliferation Research and Development. The project’s mission was to take the first steps toward developing a demonstration prototype system that uses multiple machine learning and data analytics methods on largescale open data sources to identify new, developing, and/or undeclared nuclear programs. Given the SRNL team’s on-site perspective of events culminating in the DOE’s decision to pursue the Savannah River Plutonium Processing Facility (SRPPF), the team targeted the identification of events and indicators in retrospective datasets that pointed to the activity of “fissile core fabrication at the Savannah River Site” prior to the official announcement in May of 2018. A preliminary modeling pipeline was developed in FY20 that showed the datasets contained adequate signal for continuation of efforts. In FY21, a modular demonstration prototype modeling pipeline has continued in development for two text-based data sources: a broad internet archive (Webhose Ltd.) and a decahose Twitter database (i.e., a global sampling of one in every ten Tweets). The techniques that have been developed rely on graph theory and anomaly detection to identify contextual shifts in key words and phrases at various points in time such that indicators of events of interest could be identified and subsequently, events could be extracted from the corpuses. The foundational concept behind the approaches is that contextual shifts in key words and phrases can act as indicators of events of interest. Both datasets have proven successful in extracting events of interest related to pit production at the Savannah River Site prior to the official announcement. In addition, the pipelines have generated a wide range of events broadly summarized as: the awarding of DOE contracts at major sites, DOE investments in various programs, accidents at DOE national laboratories, speculations about the fate of pit production in the DOE complex, domestic and international shipments and receipts of nuclear materials at DOE sites, termination of non-proliferation agreements with Russia, termination of MOX, new weapons development approvals/testing, nuclear posture reviews, major DOE cleanup/production milestones, political opinions, and nuclear watch groups’ opinions, among many others.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Machine Reading at Scale: A Search Engine for Scientific and Academic Research

The Internet, much like our universe, is ever-expanding. Information, in the most varied formats, is continuously added to the point of information overload. Consequently, the ability to navigate this ocean of data is crucial in our day-to-day lives, with familiar tools such as search engines carving a path through this unknown. In the research world, articles on a myriad of topics with distinct complexity levels are published daily, requiring specialized tools to facilitate the access and assessment of the information within. Recent endeavors in artificial intelligence, and in natural language processing in particular, can be seen as potential solutions for breaking information overload and provide enhanced search mechanisms by means of advanced algorithms. As the advent of transformer-based language models contributed to a more comprehensive analysis of both text-encoded intents and true document semantic meaning, there is simultaneously a need for additional computational resources. Information retrieval methods can act as low-complexity, yet reliable, filters to feed heavier algorithms, thus reducing computational requirements substantially. In this work, a new search engine is proposed, addressing machine reading at scale in the context of scientific and academic research. It combines state-of-the-art algorithms for information retrieval and reading comprehension tasks to extract meaningful answers from a corpus of scientific documents. The solution is then tested on two current and relevant topics, cybersecurity and energy, proving that the system is able to perform under distinct knowledge domains while achieving competent performance.

Sousa, Norberto (ORCID:0000000329194817)↗
DOE OSTI2023

Natural Language Processing for Text Based Event Extraction: Identifying Events of Interest Related to Worldwide State-Sponsored Civil Nuclear Power

Beginning in FY20, SRNL was funded by the National Nuclear Security Administration’s Office of Defense Nuclear Non-Proliferation Research and Development to develop a prototype natural language processing/natural language understating machine learning-based modeling and analysis pipeline to extract and forecast events of interest from massive open data sources. The working hypothesis within the approach is that contextual shifts in key words and phrases act as indicators of events of interest over time. Therefore, by identifying points in time where contextual shifts occur, events of interest can be extracted along with explicit and implicit connections of entities and activities. The development of the preliminary prototype pipeline proved successful, meriting further testing of the pipeline on more broad topical domains and in a worldwide data environment. Therefore, SRNL, in collaboration with the Sanghani Center for Artificial Intelligence and Data Analytics at Virginia Tech, have continued development with a test case of identifying events of interest related to worldwide state-sponsored civil nuclear power in open data sources. In the first year of this follow-on effort, the team has curated domain-specific data corpuses using an automated scheme and applied the modeling and analysis pipeline. This robust, focused, and efficient approach consists of an ensemble of analyses applied to time dependent word embedding models that are trained on the data corpuses. In this report, the team has demonstrated the capability of the existing pipeline (as development has continued in parallel) by exploring several specific case-studies centered around Rosatom’s international activities regarding the planning, construction, operation, and/or shutdown of nuclear reactors. A basic timeline events has been generated by manually cataloging known “milestone” events that have occurred at reactors in Turkey, Finland, Hungary, and Egypt and compared with the output of the modeling pipeline. In this approach, the team has characterized the lead time using the prototype pipeline, as well as the ability to capture relevant information, which proved 100% successful. A deep dive example of the Akkuyu reactor (Turkey) is presented that shows the breadth of information that can be captured using the approach. In this case study, events were extracted pertaining to the planning/construction of Akkuyu including protests from the population, information campaigns in response to the protests, forged regulatory documents and lawsuits, budgetary/shareholder information, geopolitical tensions, and the various construction milestones. This has demonstrated the pipeline’s utility as a research aid or real-time event extraction tool, where summary-level information and detailed text extractions from millions of articles or Tweets across long time periods can be generated with significantly less effort than current techniques.

22 GENERAL STUDIES OF NUCLEAR REACTORS↗
DOE OSTI2025

MechBERT: Language Models for Extracting Chemical and Property Relationships about Mechanical Stress and Strain

Language models are transforming materials-aware naturallanguage processing by enabling the extraction of dynamic, context-rich information from unstructured text, thus, moving beyond the limitations of traditional information-extraction methods. Moreover, small language models are on the rise because some of them can perform better than large language models (LLMs) when given domain-specific questionanswer tasks, especially about an application area that relies on a highly specialized vernacular, such as materials science. We therefore present a new class of MechBERT language models for understanding mechanical stress and strain in materials. These employ Bidirectional Encoder Representations for transformer (BERT) architectures. We showcase four MechBERT models, all of which were pretrained on a corpus of documents that are textually rich in chemicals and their stress–strain properties and were fine-tuned on question-answering tasks. We evaluated the level of performance of our models on domain-specific as well as general English-language question-answer tasks and also explored the influence of the size and type of BERT architectures on model performance. We find that our MechBERT models outperform BERT-based models of the same size and maintain relevancy better than much larger BERT-based models when tasked with domain-specific question-answering tasks within the stress–strain engineering sector. These small language models also enable much faster processing and require a much smaller fraction of data to pretrain them, affording them greater operational efficiency and energy sustainability than LLMs.

37 INORGANIC, ORGANIC, PHYSICAL, AND ANALYTICAL CH↗
DOE OSTI2021

Artificial Judgement Assistance from teXt (AJAX): Applying Open Domain Question Answering to Nuclear Non-proliferation Analysis

Nuclear non-proliferation analysis is complex and subjective, as the data is sparse, and examples are rare and diverse. While analysing non-proliferation data, it is often desired that the findings be completely auditable such that any claim or assertion can be sourced directly to the reference material from which it was derived. Currently this is accomplished by analysts thoroughly documenting underlying assumptions and clearly referencing details to source documents. This is a labour-intensive and time-consuming process that can be difficult to scale with geometrically increasing quantities of data. In this work, we describe an approach to leverage bi-directional language models for nuclear non-proliferation analysis. It has been shown recently that these models not only capture language syntax but also some of the relational knowledge present in the training data. We have devised a unique Salt and Pepper strategy for testing the knowledge present in the language models, while also introducing auditability function in our pipeline. We demonstrate that fine-tuning the bi-directional language models on domain specific corpus improves their ability to answer domain-specific factoid questions. Our hope is that the results presented in this paper will further the natural language processing (NLP) field by introducing the ability to audit the answers provided by the language models to bring forward the source of said knowledge.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗
DOE OSTI2023

K-RET: knowledgeable biomedical relation extraction system

Abstract Motivation Relation extraction (RE) is a crucial process to deal with the amount of text published daily, e.g. to find missing associations in a database. RE is a text mining task for which the state-of-the-art approaches use bidirectional encoders, namely, BERT. However, state-of-the-art performance may be limited by the lack of efficient external knowledge injection approaches, with a larger impact in the biomedical area given the widespread usage and high quality of biomedical ontologies. This knowledge can propel these systems forward by aiding them in predicting more explainable biomedical associations. With this in mind, we developed K-RET, a novel, knowledgeable biomedical RE system that, for the first time, injects knowledge by handling different types of associations, multiple sources and where to apply it, and multi-token entities. Results We tested K-RET on three independent and open-access corpora (DDI, BC5CDR, and PGR) using four biomedical ontologies handling different entities. K-RET improved state-of-the-art results by 2.68% on average, with the DDI Corpus yielding the most significant boost in performance, from 79.30% to 87.19% in F-measure, representing a P-value of 2.91×10−12. Availability and implementation https://github.com/lasigeBioTM/K-RET.

Sousa, Diana F. (ORCID:0000000305979273)↗
DOE OSTI2024

Development of message passing-based graph convolutional networks for classifying cancer pathology reports

Abstract Background Applying graph convolutional networks (GCN) to the classification of free-form natural language texts leveraged by graph-of-words features (TextGCN) was studied and confirmed to be an effective means of describing complex natural language texts. However, the text classification models based on the TextGCN possess weaknesses in terms of memory consumption and model dissemination and distribution. In this paper, we present a fast message passing network (FastMPN), implementing a GCN with message passing architecture that provides versatility and flexibility by allowing trainable node embedding and edge weights, helping the GCN model find the better solution. We applied the FastMPN model to the task of clinical information extraction from cancer pathology reports, extracting the following six properties: main site, subsite, laterality, histology, behavior, and grade. Results We evaluated the clinical task performance of the FastMPN models in terms of micro- and macro-averaged F1 scores. A comparison was performed with the multi-task convolutional neural network (MT-CNN) model. Results show that the FastMPN model is equivalent to or better than the MT-CNN. Conclusions Our implementation revealed that our FastMPN model, which is based on the PyTorch platform, can train a large corpus (667,290 training samples) with 202,373 unique words in less than 3 minutes per epoch using one NVIDIA V100 hardware accelerator. Our experiments demonstrated that using this implementation, the clinical task performance scores of information extraction related to tumors from cancer pathology reports were highly competitive.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2023

Linguistic variation in functional types of statutory law

When the meaning of an ambiguous word, phrase or grammatical structure in a statutory provision is disputed, courts are tasked with identifying the best meaning of the contested language. A common method of resolving linguistic ambiguities is to investigate the meaning of the contested word or structure in statutory provisions with similar subject matter. While the subject matter of a text has a demonstrated effect on language use, register variation research shows that the function of a text is also highly influential in predicting linguistic variation. Thus far, the function of a statutory provision (e.g., obligation to act, authorization to act) has not been considered in legal interpretative research. In the present study, I investigate the extent to which function influences the lexico-grammatical characteristics of statutory texts. 2,573 statutory provisions from the Arizona State Code are individually assigned to one of seven categories representing their function: Duties, Permissions, Impersonal Rules, Operational Definitions, Prohibitions, Procedural Guidelines, and Criminal Offenses. Key feature analysis is used to identify and describe patterns of lexico-grammatical variation between the seven functional types. Results reveal a great deal of lexico-grammatical variation associated with function in the register of statutory law. Furthermore, some functional types of statutory provisions are more linguistically distinct than others. In conclusion, these findings suggest that it may be beneficial to consider communicative function when investigating legal interpretative questions.

99 GENERAL AND MISCELLANEOUS↗
DOE OSTI2025

AI Model Benchmarking for Nonproliferation Applications: Steel Thread Benchmarking Task Force Technical Report (Rev. 2)

Steel Thread is a NA-22 venture that seeks to build trustworthy, reliable AI models that can be used in a wide variety of nonproliferation tasks. A key aspect of building these models is developing appropriate benchmarks and evaluation methods, which will enable the venture to identify and adapt models to provide the most value in the nonproliferation domain. Benchmarks must be relevant to key tasks in this domain, such as question answering, information retrieval, document summarization and classification, consensus analysis, and image and data analysis. This report 1) provides an overview of benchmark design, evaluation, and challenges; 2) reviews a variety of open benchmarks, with a focus on language models and tasks; and 3) identifies benchmarks that are most relevant to Steel Thread. This report is intended to serve as a basis for further efforts to classify and evaluate benchmarks and their correlation with success on nonproliferation-specific tasks. The Steel Thread venture has defined benchmarks to be a particular combination of a dataset (or datasets) and a metric (or metrics) conceptualized as representing one or more specific tasks or sets of abilities for a specific modality. It is adopted by a research community as a shared framework for comparing methods.1 It includes 1) Data: Labeled (a designated subset not used for training, which could be all the data), 2) Metric: A way to quantify performance, 3) Task/Ability: What the benchmark is testing, 4) Protocol: A structured and repeatable evaluation process, 5) Baseline/Reference Model: For comparison; could be statistical, rule-based, SME-derived, or another model, and 6) Maintenance Plan: to update with new information over time; important for long-term utility. For further clarity, the definition includes what a benchmark, in this context, is not. It is not a corpus of training data, specific to a model (it is intended to apply to a range of models), a universal evaluation of performance, a guarantee that the ‘top’ model on the leaderboard will be the best fit for every specific use case, an all-encompassing proof of a model’s universal quality, nor is it a one-size-fits-all measure of success. It does not cover every real-world constraint (like operational, ethical, or cost considerations), a systems integration test, or a unit test. This definition was inspired by and resulted from discussions within the Steel Thread Benchmarking Task Force. This group was formed to define what we would mean as a benchmark within Steel Thread but persisted as the need to develop a thorough understanding of the large and expanding existing benchmarking space. This technical report is a result of the group’s divide and conquer approach to exploring this space. The release of benchmarks might not be progressing as quickly as model development, but it is moving very fast, as many benchmarks quickly become saturated, when state-of-the-art models score so close to the benchmark’s ceiling that their results are virtually indistinguishable. At that point, the test no longer differentiates between new systems, so researchers usually stop reporting scores as the benchmark no longer informs about improvements from the next generation of models. In the OpenAI announcement of GPT-5, they reported results on six flagship public benchmarks (AIME 2025, SWE-bench Verified, Aider Polyglot, MMMU, HealthBench Hard, GPQA) but the full system-card covers roughly thirty-five separate evaluations, comprising hundreds of test task items in total. There have been some efforts to summarize benchmarks in specific fields, like for text-to-image generation, but these surveys have had a narrow methodology scope. Therefore, a comprehensive survey of all benchmarks or even all benchmarks that could be relevant to Steel Thread is outside of the scope of this report. We chose some specific benchmarks to investigate in detail.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2023

Large Language Models (LLMs) for Energy Systems Research

The integration of Large Language Models (LLMs) in energy systems research promises transformative results, as demonstrated in this work, particularly in the realms of information retrieval and legal document analysis. We have developed a chat-based interface, specifically designed to query an extensive corpus of technical reports from the National Renewable Energy Laboratory (NREL). This interface capitalizes on the natural language processing capabilities of LLMs, providing future consumers of NREL research with a user-friendly platform to access and extract valuable information from technical documents, thus enhancing the dissemination of research to the public. In addition to information retrieval, we have employed LLMs to extract renewable energy siting ordinances from a variety of legal documents, a task traditionally driven by significant human labor. This automated extraction not only supports the ongoing development of the high-impact NREL siting ordinance database but also ensures the database's accuracy and comprehensiveness. Crucially, we have augmented the performance of LLMs through the integration of a decision tree framework, resulting in a substantial improvement in extraction accuracy. Comparative analysis with manual efforts has shown that this approach not only rivals but also significantly surpasses human accuracy, heralding increased reliability in legal document analysis for energy systems research. To democratize access to these advancements and foster collaborative research, we introduce the "Energy Language Model" (ELM), an open-source software package. ELM encapsulates the methodologies and tools developed in this work, providing researchers and practitioners with a robust toolkit to conduct similar analyses within their respective domains. Through these contributions, this work underscores the immense potential of LLMs in revolutionizing energy systems research, improving accuracy, efficiency, and accessibility in the field.

automation↗