arXiv subjects

Lexington Whalen

Publications and source records attributed to Lexington Whalen.

8 recordsLinked to original sources

arXiv2026

KBBQ: A Predictive Noise Law and the Limits of Spectrum Flattening in FP4 Quantization

We develop a second-order theory of quantization noise in matrix multiplication in which the quantization format is characterized by the variance it assigns to each element. The constant variance profile of integer quantization recovers existing integer-noise theory, while the multiplicative profile of floating-point rounding reduces the data dependence to a scalar, the participation factor $κ$, yielding a closed-form signal-to-noise-ratio law. The resulting functional also admits a closed-form upper bound $κ^{*}$ that no function-preserving linear transform can exceed and that is attained by a recent state-of-the-art method. Building on this analysis, we introduce KBBQ (\textbf{K}appa-\textbf{B}raked \textbf{B}lockwise \textbf{Q}uantization), which parameterizes the extent to which a transform approaches this ceiling. At W4A4, across four base models and two FP4 formats, KBBQ outperforms the prior state of the art without additional deployment-time computation.

cs.LG↗

arXiv2026

Nemotron-Labs-Diffusion: A Tri-Mode Language Model Unifying Autoregressive, Diffusion, and Self-Speculation Decoding

We introduce Nemotron-Labs-Diffusion, a tri-mode language model (LM) that unifies AR, diffusion, and self-speculation decoding within a single architecture. Trained with a joint AR-diffusion objective, Nemotron-Labs-Diffusion can switch modes to sustain high throughput across deployment settings and concurrency levels. Our study shows that (1) AR and diffusion objectives are complementary: diffusion improves lookahead planning, while AR provides left-to-right linguistic priors. (2) In self-speculation mode, diffusion drafts while AR verifies, outperforming multi-token prediction (MTP) methods in both acceptance rate and real-device efficiency. (3) A speed-of-light analysis further demonstrates diffusion's long-term potential, with up to 76.5% more tokens per forward pass than self-speculation under an optimal sampler. Scaling to 3B, 8B, and 14B parameters, our Nemotron-Labs-Diffusion family, including base, instruct, and vision-language models, consistently outperforms state-of-the-art open-source AR and diffusion LMs in both accuracy and speed. For example, Nemotron-Labs-Diffusion-8B decodes 6x more tokens per forward than Qwen3-8B with comparable accuracy, translating to 4x higher throughput on SPEED-Bench with SGLang on a GB200 GPU.

cs.CL↗

arXiv2026

Teaching Diffusion to Speculate Left-to-Right

Large language models (LLMs) achieve remarkable performance across a wide range of tasks, but their autoregressive decoding process incurs substantial inference costs due to inherently sequential token generation. Speculative decoding addresses this bottleneck by employing a lightweight draft model to propose multiple future tokens that are subsequently verified in parallel by a larger target model. Recent work has demonstrated that diffusion language models are well suited for this setting, as they can generate entire blocks of draft tokens in parallel and thereby alleviate the sequential constraints of autoregressive drafting. A subtlety of this regime is that block-diffusion drafters generate tokens bidirectionally within a block, whereas verification is performed by an autoregressive target model that evaluates tokens in a strictly left-to-right manner, leaving a gap between the symmetric training-time objective and the asymmetric verification-time reward. In this work, we offer an empirical analysis of three training-time interventions that narrow this gap: token positional weighting, a first-error focal loss that targets the position that breaks the accepted prefix within each block, and a chain loss term that substitutes a differentiable surrogate for the expected accepted length. The three interventions act along orthogonal axes (position, block-conditional first error, joint prefix) and compose additively; they are likewise orthogonal to test-time alignment mechanisms such as multi-draft self-selection, with which they can in principle be combined. Across four target models and six reasoning, code, and dialogue benchmarks, the three interventions raise accepted draft length by 21-76% per benchmark over a position-uniform baseline, without adding additional forward passes and without changing the inference pipeline or the rejection-sampling exactness contract.

cs.CL↗

arXiv2026

Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speed

Diffusion language models (dLMs) have emerged as a promising paradigm that enables parallel, non-autoregressive generation, but their learning efficiency lags behind that of autoregressive (AR) language models when trained from scratch. To this end, we study AR-to-dLM conversion to transform pretrained AR models into efficient dLMs that excel in speed while preserving AR models' task accuracy. We achieve this by identifying limitations in the attention patterns and objectives of existing AR-to-dLM methods and then proposing principles and methodologies for more effective AR-to-dLM conversion. Specifically, we first systematically compare different attention patterns and find that maintaining pretrained AR weight distributions is critical for effective AR-to-dLM conversion. As such, we introduce a continuous pretraining scheme with a block-wise attention pattern, which remains causal across blocks while enabling bidirectional modeling within each block. We find that this approach can better preserve pretrained AR models' weight distributions than fully bidirectional modeling, in addition to its known benefit of enabling KV caching, and leads to a win-win in accuracy and efficiency. Second, to mitigate the training-test gap in mask token distributions (uniform vs. highly left-to-right), we propose a position-dependent token masking strategy that assigns higher masking probabilities to later tokens during training to better mimic test-time behavior. Leveraging this framework, we conduct extensive studies of dLMs' attention patterns, training dynamics, and other design choices, providing actionable insights into scalable AR-to-dLM conversion. These studies lead to the Efficient-DLM family, which outperforms state-of-the-art AR models and dLMs, e.g., our Efficient-DLM 8B achieves +5.4%/+2.7% higher accuracy with 4.5x/2.7x higher throughput compared to Dream 7B and Qwen3 4B, respectively.

cs.CL↗

arXiv2025

Early-Bird Diffusion: Investigating and Leveraging Timestep-Aware Early-Bird Tickets in Diffusion Models for Efficient Training

Training diffusion models (DMs) requires substantial computational resources due to multiple forward and backward passes across numerous timesteps, motivating research into efficient training techniques. In this paper, we propose EB-Diff-Train, a new efficient DM training approach that is orthogonal to other methods of accelerating DM training, by investigating and leveraging Early-Bird (EB) tickets -- sparse subnetworks that manifest early in the training process and maintain high generation quality. We first investigate the existence of traditional EB tickets in DMs, enabling competitive generation quality without fully training a dense model. Then, we delve into the concept of diffusion-dedicated EB tickets, drawing on insights from varying importance of different timestep regions. These tickets adapt their sparsity levels according to the importance of corresponding timestep regions, allowing for aggressive sparsity during non-critical regions while conserving computational resources for crucial timestep regions. Building on this, we develop an efficient DM training technique that derives timestep-aware EB tickets, trains them in parallel, and combines them during inference for image generation. Extensive experiments validate the existence of both traditional and timestep-aware EB tickets, as well as the effectiveness of our proposed EB-Diff-Train method. This approach can significantly reduce training time both spatially and temporally -- achieving 2.9$\times$ to 5.8$\times$ speedups over training unpruned dense models, and up to 10.3$\times$ faster training compared to standard train-prune-finetune pipelines -- without compromising generative quality. Our code is available at https://github.com/GATECH-EIC/Early-Bird-Diffusion.

cs.CV↗

arXiv2024

DataDock: An Open Source Data Hub for Research

Every research project necessitates data, often requiring sharing and collaborative review within a team. However, there is a dearth of good open-source data sharing and reviewing services. Existing file-sharing services generally mandate paid subscriptions for increased storage or additional members, diverting research funds from addressing the core research problem that a lab is attempting to work on. Moreover, these services often lack direct features for reviewing or commenting on data quality, a vital part of ensuring high quality data generation. In response to these challenges, we present DataDock, a specialized file transfer service crafted for specifically for researchers. DataDock operates as an application hosted on a research lab server. This design ensures that, with access to a machine and an internet connection, teams can facilitate file storage, transfer, and review without incurring extra costs. Being an open-source project, DataDock can be customized to suit the unique requirements of any research team, and is able to evolve to meet the needs of the research community. We also note that there are no limitations with respect to what data can be shared, downloaded, or commented on. As DataDock is agnostic to the file type, it can be used in any field from bioinformatics to particle physics; as long as it can be stored in a file, it can be shared. We open source the code here: https://github.com/lxaw/DataDock

cs.DB↗

arXiv2023

Wordification: A New Way of Teaching English Spelling Patterns

Literacy, or the ability to read and write, is a crucial indicator of success in life and greater society. It is estimated that 85% of people in juvenile delinquent systems cannot adequately read or write, that more than half of those with substance abuse issues have complications in reading or writing and that two-thirds of those who do not complete high school lack proper literacy skills. Furthermore, young children who do not possess reading skills matching grade level by the fourth grade are approximately 80% likely to not catch up at all. Many may believe that in a developed country such as the United States, literacy fails to be an issue; however, this is a dangerous misunderstanding. Globally an estimated 1.19 trillion dollars are lost every year due to issues in literacy; in the USA, the loss is an estimated 300 billion. To put it in more shocking terms, one in five American adults still fail to comprehend basic sentences. Making matters worse, the only tools available now to correct a lack of reading and writing ability are found in expensive tutoring or other programs that oftentimes fail to be able to reach the required audience. In this paper, our team puts forward a new way of teaching English spelling and word recognitions to grade school students in the United States: Wordification. Wordification is a web application designed to teach English literacy using principles of linguistics applied to the orthographic and phonological properties of words in a manner not fully utilized previously in any computer-based teaching application.

cs.OH↗

arXiv2023

On Creating a Comprehensive Food Database

Studies with the primary aim of addressing eating disorders focus on assessing the nutrient content of food items with an exclusive focus on caloric intake. There are two primary impediments that can be noted in these studies. The first of these relates to the fact that caloric intake of each food item is calculated from an existing database. The second concerns the scientific significance of caloric intake used as the single measure of nutrient content. By requiring an existing database, researchers are forced to find some source of a comprehensive set of food items as well as their respective nutrients. This search alone is a difficult task, and if completed often leads to the requirement of a paid API service. These services are expensive and non-customizable, taking away funding that could be aimed at other parts of the study only to give an unwieldy database that can not be modified or contributed to. In this work, we introduce a new rendition of the USDA's food database that includes both foods found in grocery stores and those found in restaurants or fast food places. At the moment, we have accumulated roughly 1.5 million food entries consisting of approximately 18,000 brands and 100 restaurants in the United States. These foods also have an abundance of nutrient data associated with them, from the caloric amount to saturated fat levels. The data is stored in MySQL format and is spread among five major tables. We have also procured images for theses foods entries when available, and have included all of our data and program scripts in an open source repository.

cs.DB↗