Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “tokenization”

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 37 records · Page 2
All sourcesNASA NTRSDOE OSTI
DOE OSTI2022

Security Evaluation of Smart Cards and Secure Tokens: Benefits and Drawbacks for Reducing Supply Chain Risks of Nuclear Power Plants

The supply chain attack pathway is being increasingly used by adversaries to bypass security controls and gain unauthorized access to sensitive networks and equipment (e.g., Critical Digital Assets). Cyber-attacks targeting supply chain generally aim to compromise the environments, products, or services of vendors and suppliers to inject, add, or substitute authentic software and hardware with malicious elements. These malicious elements are deemed to be authentic as they arise from the vendor or supplier (i.e., the supply chain). This research aims to leverage findings and assumptions made from the previous report to determine the security benefits and drawbacks of a smart card- based hardware root of trust. Smart cards can provide devices inside Nuclear Power Plants (NPP) with a secure environment to store keys in and perform sensitive operations such as digital signature generation. These abilities can be leveraged to increase supply chain cybersecurity by autonomously providing NPP Licensees with reports on device integrity, authenticity and measurements of executable and non-executable data.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗
DOE OSTI2022

Name Normalizer

Our name normalizer and enrichment micro-service address the shortcomings of standard text normalization. Initially the text input is split into tokens. This process takes into account some conventions of formatting. Any dashes present between two tokens preserves the relationship of those two tokens. Conversely a comma between two tokens ensures that the separation between the two tokens is maintained. The order of tokens is also preserved. Standard text normalization (conversion to lowercase, trimming extra white space, and canonicalization) is then applied to the tokens. Once normalized every unique and sequence of tokens is given confidence values by comparing the normalized values to publicly available data.

Bleeker, Amelia↗
DOE OSTI2026

MTRE

Multi-Token Reliability Estimation (MTRE) is a lightweight, white-box hallucination detector for vision-language models. Instead of using only the first output token, MTRE aggregates logits from the first ~10 tokens and feeds them to a small attention-based reliability head; per-token scores are combined via a sequential log-likelihood-ratio test with early-stopping, and an MTRE-t variant calibrates thresholds via cross-fitting. MTRE reports average gains of +9.4% Accuracy and +14.8% AUROC over common baselines across MAD-Bench, MM-SafetyBench, MathVista, and arithmetic/counting tasks, while adding ~4.3M params and ~1% inference overhead (~26 MB VRAM, ~0.94 ms per detection). Key limitation: requires access to early token logits and is evaluated on a handful of open-source 7B VLMs.

Bhattarai, Manish [Los Alamos National Labs]↗
DOE OSTI2026

MATEY: multiscale adaptive transformer models for spatiotemporal physical systems

Accurate representation of the multiscale features in spatiotemporal physical systems using vision transformer architectures requires extremely long, computationally prohibitive token sequences. To address this issue, we propose two novel adaptive tokenization schemes that dynamically adjust patch sizes based on local features: one ensures convergent behavior to uniform patch refinement, while the other offers better computational efficiency. Moreover, we present a set of spatiotemporal attention schemes, where the temporal or axial spatial dimensions are decoupled, to evaluate their baseline computational and data efficiencies and to determine whether adaptive tokenization can improve this performance. We assess the performance of the proposed multiscale adaptive model, MATEY, in a sequence of experiments. Compared to a full spatiotemporal attention scheme or a scheme that decouples only the temporal dimension, we find that fully decoupled axial attention is less efficient and expressive, requiring more training time and model parameters to achieve the same accuracy. The experiments on the adaptive tokenization schemes show that, compared to a uniformly refined model, the proposed schemes achieve comparable or improved accuracy at a much lower cost in the tested two-dimensional settings. While the asymptotic analysis suggests the potential for favorable scaling, empirical validation at substantially longer sequence lengths remains to be performed in future work. Finally, we demonstrate in two fine-tuning tasks featuring different physics that models pretrained on PDEBench data outperform the ones trained from scratch, especially in the low data regime with frozen attention.

adaptive tokenization↗
DOE OSTI2025

Debiasing Watermarks for Large Language Models via Maximal Coupling

Watermarking language models is essential for distinguishing between human and machine-generated text and thus maintaining the integrity and trustworthiness of digital communication. Here, we present a novel green/red list watermarking approach that partitions the token set into “green” and “red” lists, subtly increasing the generation probability for green tokens. To correct token distribution bias, our method employs maximal coupling, using a uniform coin flip to decide whether to apply bias correction, with the result embedded as a pseudorandom watermark signal. Theoretical analysis confirms this approach’s unbiased nature and robust detection capabilities. Experimental results show that it outperforms prior techniques by preserving text quality while maintaining high detectability, and it demonstrates resilience to targeted modifications aimed at improving text quality. This research provides a promising watermarking solution for language models, balancing effective detection with minimal impact on text quality.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

Position-Enhanced Gradient Attack (PEGA) on Medical Language Models

Federated Learning (FL) enables collaborative training of language models on sensitive clinical notes without sharing the data. However, this paradigm is vulnerable to gradient inversion attacks that can reconstruct private data from shared gradients. We find that state-of-the-art attacks are less effective in the medical domain, failing to overcome the unique challenges posed by its specialized vocabulary and unstructured format. To address this, we introduce the Position-Enhanced Gradient Attack (PEGA), a novel attack that makes gradients position-aware by optimizing token and position embeddings simultaneously. PEGA employs two key innovations: a periodic sorting of positional embeddings to resolve token order ambiguity and a late-stage embedding replacement strategy to correct hard-to-recover critical tokens. To evaluate the leakage of sensitive data more directly, we also propose the Unified PHI-Recall (UPHI), a new metric measuring the recovery of Protected Health Information. Experiments on the MIMIC-III dataset show that PEGA significantly outperforms leading attacks like TAG and LAMP, particularly in its ability to reconstruct identifiable patient information, exposing a more severe and nuanced privacy risk in federated medical NLP.

Xu, Nuo [University of Minnesota]↗
DOE OSTI2026

XMark: Reliable Multi-Bit Watermarking for LLM-Generated Texts

Multi-bit watermarking has emerged as a promising solution for embedding imperceptible binary messages into Large Language Model (LLM)-generated text, enabling reliable attribution and tracing of malicious usage of LLMs. Despite recent progress, existing methods still face key limitations: some become computationally infeasible for large messages, while others suffer from a poor trade-off between text quality and decoding accuracy. Moreover, the decoding accuracy of existing methods drops significantly when the number of tokens in the generated text is limited, a condition that frequently arises in practical usage. To address these challenges, we propose XMark, a novel method for encoding and decoding binary messages in LLM-generated texts. The unique design of XMark’s encoder produces a less distorted logit distribution for watermarked token generation, preserving text quality, and also enables its tailored decoder to reliably recover the encoded message with limited tokens. Extensive experiments across diverse downstream tasks show that XMark significantly improves decoding accuracy while preserving the quality of watermarked text, outperforming prior methods. The code will be made publicly available upon acceptance.

Xu, Jiahao [University of Nevada, Reno]↗
DOE OSTI2026

Neural Scaling Laws for Jet Generation

Recently observed empirical scaling laws describe the performance of foundation-type models as three independent key quantities -- dataset size, compute, and model parameters -- are modified. Extracting these scaling laws informs the training of large complex models for which the tuning of hyperparameters in traditional ways is not feasible. This work for the first time explores if scaling laws can also be observed for the task of particle jet generation -- both relevant as a pre-training objective for foundation models and as in-situ simulation by itself. We indeed replicate the key logarithmic scaling law behavior for model-size scaling. Beyond studying the next token prediction validation loss of the generative model, we also study the sliced Wasserstein distance of five physical quantities that are not immediately available to the model during training. Our study shows that this quantity is monotonically related to the next token prediction validation loss, meaning that this loss is indeed a good proxy for the physics performance. For the scaling with dataset size and compute, we observe substantially weaker scaling behavior of both the loss and the sliced Wasserstein distance. We analyze this behavior by introducing the concept of a learnable window, and argue that autoregressive next token prediction on jet constituents exhibits comparatively rapid saturation relative to language-model studies. We discuss possible origins of this behavior, including the stochastic nature of QCD radiation and differences between generative and supervised learning tasks in collider physics.

Amram, Oz [Fermilab]↗
DOE OSTI2022

Automating Genetic Algorithm Mutations for Molecules Using a Masked Language Model

Inspired by the evolution of biological systems, genetic algorithms have been applied to generate solutions for optimization problems in a variety of scientific and engineering disciplines. For a given problem, a suitable genome representation must be defined along with a mutation operator to generate subsequent generations. Unlike natural systems which display a variety of complex rearrangements (e.g. mobile genetic elements), mutation for genetic algorithms commonly utilizes only random point-wise changes. Furthermore, generalizing beyond point-wise mutations poses a key difficulty as useful genome rearrangements depend on the representation and problem domain. To move beyond the limitations of manually defined point-wise changes, here we propose the use of techniques from masked language models to automatically generate mutations. As a first step, common subsequences within a given population are used to generate a vocabulary. The vocabulary is then used to tokenize each genome. A masked language model is trained on the tokenized data in order to generate possible rearrangements (i.e. mutations). In order to illustrate the proposed strategy, we use string representations of molecules and use a genetic algorithm to optimize for drug-likeness and synthesizability. Finally, our results show that moving beyond random point-wise mutations accelerates genetic algorithm optimization.

59 BASIC BIOLOGICAL SCIENCES↗
DOE OSTI2024

EHR-BERT: A BERT-based model for effective anomaly detection in electronic health records

Objective: Physicians and clinicians rely on data contained in electronic health records (EHRs), as recorded by health information technology (HIT), to make informed decisions about their patients. The reliability of HIT systems in this regard is critical to patient safety. Consequently, better tools are needed to monitor the performance of HIT systems for potential hazards that could compromise the collected EHRs, which in turn could affect patient safety. In this paper, we propose a new framework for detecting anomalies in EHRs using sequence of clinical events. This new framework, EHR-Bidirectional Encoder Representations from Transformers (BERT), is motivated by the gaps in the existing deep-learning related methods, including high false negatives, sub-optimal accuracy, higher computational cost, and the risk of information loss. EHR-BERT is an innovative framework rooted in the BERT architecture, meticulously tailored to navigate the hurdles in the contemporary BERT method; thus, enhancing anomaly detection in EHRs for healthcare applications.Methods: The EHR-BERT framework was designed using the Sequential Masked Token Prediction (SMTP) method. This approach treats EHRs as natural language sentences and iteratively masks input tokens during both training and prediction stages. This method facilitates the learning of EHR sequence patterns in both directions for each event and identifies anomalies based on deviations from the normal execution models trained on EHR sequences.Results: Extensive experiments on large EHR datasets across various medical domains demonstrate that EHR-BERT markedly improves upon existing models. It significantly reduces the number of false positives and enhances the detection rate, thus bolstering the reliability of anomaly detection in electronic health records. This improvement is attributed to the model’s ability to minimize information loss and maximize data utilization effectively.Conclusion: EHR-BERT showcases immense potential in decreasing medical errors related to anomalous clinical events, positioning itself as an indispensable asset for enhancing patient safety and the overall standard of healthcare services. The framework effectively overcomes the drawbacks of earlier models, making it a promising solution for healthcare professionals to ensure the reliability and quality of health data.

96 KNOWLEDGE MANAGEMENT AND PRESERVATION↗
DOE OSTI2026

Structural and compositional complexities of hierarchical self-assembly: A hypergraph approach

Programmable self-assembly enables the construction of complex molecular, supramolecular, and crystalline architectures from well-designed building blocks. In this work, we introduce a hypergraph-based formalism, Blocks & Bonds (B&B), which generalizes classical chemical graph theory by incorporating directed and multicolored interactions, internal symmetries, and hierarchical organization. Within this framework, we develop the Structure Code (SC), a compact and versatile language for describing self-assembled architectures. We define a Kolmogorov-style structural complexity as the total information content of SC, obtained through its tokenization and Shannon information assignment. Complementing this encoding-based measure, we introduce a much simpler quantity, the compositional complexity, which depends only on the number and cumulative usage of block and bond types in the construction set. A central result of this work is a strong empirical correlation between the token-based structural complexity and the compositional complexity across all examined systems. Owing to this agreement, the compositional complexity emerges as the most practical and broadly applicable measure: it is easy to compute, requires no explicit encoding, and yet closely tracks the actual information content of structurally diverse architectures. Applications to molecular systems (ethylene glycol and glucose), DNA-origami lattices, and crystalline assemblies show that B&B hypergraphs provide a unified, scalable, and information-efficient representation of structural organization, naturally capturing symmetry, modularity, and stereochemistry. This framework establishes a quantitative foundation for complexity-aware classification and inverse design of programmable matter.

36 MATERIALS SCIENCE↗
DOE OSTI2025

Learning-Based Quantum Compilation: Translating QASM to QIR with CodeBERT

We propose a learning-based approach to quantum compilation by translating OpenQASM to Quantum Intermediate Representation (QIR) using a fine-tuned CodeBERT model. Trained on 10,000 synthetic QASM-QIR pairs, the model captures code semantics while addressing QIR verbosity and the 512-token limit via a custom token compression scheme. Finetuning was performed on the Frontier supercomputer, with results showing syntactic correctness and stable validation loss reduction. Our method moves toward enabling flexible, language-modeldriven quantum software tools. It also introduces syntax error handling and the possibility of incorporating classical control constructs, addressing limitations in existing rule-based compilers like qBraid-QIR. While the current model has been validated on quantum-only circuits, we propose future evaluations on hybrid quantum-classical examples. This poster will provide architecture insights, compression examples, training loss plots, and QIR outputs. Our work highlights the potential for scalable, adaptable compilation in future quantum toolchains.

Afrose, Sharmin [ORNL]↗
DOE OSTI2026

OpenStudio®-MCP [SWR-26-035]

OpenStudio®-MCP is a Model Context Protocol (MCP) server that lets AI assistants perform building energy modeling through natural language. Rather than requiring users to learn the OpenStudio® SDK, EnergyPlus® scripting, or Ruby/Python automation, the server translates conversational requests into sequences of tool calls that create models, design HVAC systems, run simulations, and extract results — all within a single chat session. The server's 124 tools are organized into a skills architecture where each skill encapsulates a domain of building energy modeling (envelope, HVAC, loads, weather, simulation, results) behind typed, LLM-friendly interfaces. High-leverage operations like applying ASHRAE 90.1 baseline systems or generating standards-compliant typical buildings are exposed as single tool calls that internally wire dozens of OpenStudio® objects. Bundled measures from ComStock™ and Openstudio® -common-measures-gem are wrapped with dedicated tools and typed arguments rather than exposed through a generic measure interface, so AI models get consistent, error-resistant recipes without needing to discover measure arguments at runtime. A key design decision is structured results extraction: six SQL-based tools return surgical ~300–1,000 token responses (end-use breakdowns, envelope summaries, HVAC sizing, timeseries data) instead of requiring the AI to parse ~100K-token raw HTML reports, making iterative design exploration practical within context window limits. The codebase is designed as a reference implementation — explicit, well-commented, and modular — so that other simulation engines (EnergyPlus® standalone, TRNSYS, DOE-2) can use it as a template for building their own MCP servers.

Ball, Brian [National Laboratory of the Rockies (N↗
DOE OSTI2025

Sequence length scaling in vision transformers for scientific images on frontier

Vision Transformers (ViTs) are pivotal for foundational models in scientific imagery, including Earth science applications, due to their capability to process large sequence lengths. While transformers for text have inspired scaling sequence lengths in ViTs, adapting these for ViTs introduces unique challenges. We develop distributed sequence parallelism for ViTs, enabling them to handle up to 1M tokens. Our approach, leveraging DeepSpeed-Ulysses and Long-Sequence-Segmentation with model sharding, is the first to apply sequence parallelism in ViT training, achieving a 94% batch scaling efficiency on 2,048 AMD-MI250X GPUs. Evaluating sequence parallelism in ViTs, particularly in models up to 10B parameters, highlighted substantial bottlenecks. We countered these with hybrid sequence, pipeline, and flash attention strategies, to scale beyond single GPU memory limits. Our method significantly enhances climate modeling accuracy by 20% in temperature predictions, marking the first training of a vision transformer model to convergence with a sequence length of 188K tokens, using full self-attention.

Tsaris, Aristeidis (aris) [ORNL] (ORCID:0000000277↗
DOE OSTI2024

Automating the Analysis of Large Language Models Responses through Zero-Shot Question Answering

Recent advancements in Large Language Models (LLMs) have shown significant potential in various applications, yet their evaluation, particularly in zero-shot question answering scenarios, remains a challenging task. In this study, our objective was to explore precision metrics for Large Language Models (LLM) and design and implement a software pipeline to automatically evaluate LLMs' outputs under zero-shot question answering. Zero-shot question answering involves a model providing answers to questions about topics it hasn't seen during training. It leverages the principles of zero-shot learning by relying on semantic understanding and generalization from related knowledge. The data used was metadata from medical databases on congenital heart disease. We explored eleven LLM metrics and selected three for our evaluation: BLEU, BERTScore, and MoverScore. BLEU calculates a score based on the overlap of n-grams (contiguous sequences of n items, typically words) between the machine-generated translation and the reference translations. Higher BLEU scores indicate better correspondence between the machine-generated and human-generated translations. BERTScore is a metric used to evaluate the quality of machine-generated text by measuring the similarity of token embeddings produced by BERT (Bidirectional Encoder Representations from Transformers) between the generated text and reference text. MoverScore is a metric that quantifies the dissimilarity between the distributions of word embeddings from machine-generated text and reference text, emphasizing semantic similarity over exact token overlap. We also introduced HBKI, a composite metric summarizing these approaches. We tested five models —GPT-3, Llama-2, Gemini 1.5 Pro, Solar 10.7B, and Mixtral-8x7b. Our software pipeline, designed and implemented using Object-Oriented Programming principles, allows users to customize the selection and extraction of features for topics of interest in their own research. Our results show that MoverScore delivered the most precise evaluation of the LLM's outputs, while Mixtral-8x7b achieved the best overall performance in extracting metadata from the databases.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2025

An Nginx-based Content Distribution Network for HEP

With the move to HTTP/WebDAV and JSON Web Tokens as a standard protocol for transfers within the WLCG distributed storage network, a large amount of off-the-shelf technologies become viable for meeting the requirements of a Storage Element (SE). In this work, we explore the capabilities and performance of the OpenResty framework, which extends the nginx server with the LuaJIT scripting language, to recreate the feature set of a SE. We demonstrate token-authenticated HTTP read, write, and WebDAV third-party copy features, as well as a storage federation with HTTP redirect, proxy, and caching capabilities. We further explore the performance scaling in terms of throughput and requests per second.

Aarora, Aashay [UC, San Diego]↗