Author name cluster

Lyle Ungar

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

15 papers

1 author row

AAAI Conference 2026 Conference Paper

LaTeX2Layout: High-Fidelity, Scalable Document Layout Annotation Pipeline for Layout Detection

Feijiang Han
Zelong Wang
Bowen Wang
Xinxin Liu
Skyler Cheung
Delip Rao
Chris Callison-Burch
Lyle Ungar

General-purpose Vision-Language Models (VLMs) are increasingly integral to modern AI systems for document understanding, yet their ability to perform fine-grained layout analysis remains severely underdeveloped. Overcoming this limitation requires large-scale, high-fidelity training datasets. However, current annotation methods that rely on parsing rendered PDFs are costly, error-prone, and difficult to scale. We propose a different paradigm: extracting ground-truth layout directly from the LaTeX compilation process rather than the final PDF. We present LaTeX2Layout, a generalizable procedural pipeline that recovers pixel-accurate bounding boxes and reading order from compiler traces. This enables the generation of a 140K-page dataset, including 120K programmatically generated synthetic variants that more than double the layout diversity of real-world data. Using this dataset, we fine-tune an efficient 3B-parameter VLM with an easy-to-hard curriculum that accelerates convergence. Our model achieves Kendall's tau=0.95 for reading order and mAP@50=0.91 for element grounding, delivering nearly 200% relative improvement over strong zero-shot baselines such as GPT-4o and Claude-3.7.

PDF Details DOI

NeurIPS Conference 2025 Conference Paper

Does Object Binding Naturally Emerge in Large Pretrained Vision Transformers?

Yihao Li
Saeed Salehi
Lyle Ungar
Konrad Kording

Object binding, the brain’s ability to bind the many features that collectively represent an object into a coherent whole, is central to human cognition. It groups low-level perceptual features into high‑level object representations, stores those objects efficiently and compositionally in memory, and supports human reasoning about individual object instances. While prior work often imposes object-centric attention (e. g. , Slot Attention) explicitly to probe these benefits, it remains unclear whether this ability naturally emerges in pre-trained Vision Transformers (ViTs). Intuitively, they could: recognizing which patches belong to the same object should be useful for downstream prediction and thus guide attention. Motivated by the quadratic nature of self-attention, we hypothesize that ViTs represent whether two patches belong to the same object, a property we term IsSameObject. We decode IsSameObject from patch embeddings across ViT layers using a quadratic similarity probe, which reaches over 90\% accuracy. Crucially, this object-binding capability emerges reliably in DINO, CLIP, and ImageNet-supervised ViTs, but is markedly weaker in MAE, suggesting that binding is not a trivial architectural artifact, but an ability acquired through specific pretraining objectives. We further discover that IsSameObject is encoded in a low-dimensional subspace on top of object features, and that this signal actively guides attention. Ablating IsSameObject from model activations degrades downstream performance and works against the learning objective, implying that emergent object binding naturally serves the pretraining objective. Our findings challenge the view that ViTs lack object binding and highlight how symbolic knowledge of “which parts belong together” emerges naturally in a connectionist system.

PDF Details

AAAI Conference 2025 Conference Paper

The Illusion of Empathy: How AI Chatbots Shape Conversation Perception

Tingting Liu
Salvatore Giorgi
Ankit Aich
Allison Lahnala
Brenda Curtis
Lyle Ungar
João Sedoc

As AI chatbots increasingly incorporate empathy, understanding user-centered perceptions of chatbot empathy and its impact on conversation quality remains essential yet under-explored. This study examines how chatbot identity and perceived empathy influence users' overall conversation experience. Analyzing 155 conversations from two datasets, we found that while GPT-based chatbots were rated significantly higher in conversational quality, they were consistently perceived as less empathetic than human conversational partners. Empathy ratings from GPT-4o annotations aligned with user ratings, reinforcing the perception of lower empathy in chatbots compared to humans. Our findings underscore the critical role of perceived empathy in shaping conversation quality, revealing that achieving high-quality human-AI interactions requires more than simply embedding empathetic language; it necessitates addressing the nuanced ways users interpret and experience empathy in conversations with chatbots.

PDF Details DOI

TMLR Journal 2023 Journal Article

Automated Detection of Causal Inference Opportunities: Regression Discontinuity Subgroup Discovery

Tony Liu
Patrick Lawlor
Lyle Ungar
Konrad Kording
Rahul Ladhania

The gold standard for the identification of causal effects are randomized controlled trials (RCT), but RCTs may not always be feasible to conduct. When treatments depend on a threshold however, such as the blood sugar threshold for diabetes diagnosis, we can still sometimes estimate causal effects with regression discontinuities (RDs). RDs are valid when units just above and below the threshold have the same distribution of covariates and thus no confounding in the presence of noise, establishing an as-if randomization. In practice however, implementing RD studies can be difficult as identifying treatment thresholds require considerable domain expertise -- furthermore, the thresholds may differ across subgroups (e.g., the blood sugar threshold for diabetes may differ across demographics), and ignoring these differences can lower statistical power. Finding the thresholds and to whom they apply is an important problem currently solved manually by domain experts, and data-driven approaches are needed when domain expertise is not sufficient. Here, we introduce Regression Discontinuity SubGroup Discovery (RDSGD), a machine-learning method that identifies statistically powerful and interpretable subgroups for RD thresholds. Using a medical claims dataset with over 60 million patients, we apply RDSGD to multiple clinical contexts and identify subgroups with increased compliance to treatment assignment thresholds. As treatment thresholds matter for many diseases and policy decisions, RDSGD can be a powerful tool for discovering new avenues for causal estimation.

PDF Details

CLeaR Conference 2022 Conference Paper

Data-driven exclusion criteria for instrumental variable studies

Tony Liu
Patrick Lawlor
Lyle Ungar
Konrad Kording

When using instrumental variables for causal inference, it is common practice to apply specific exclusion criteria to the data prior to estimation. This exclusion, critical for study design, is often done in an ad hoc manner, informed by a priori hypotheses and domain knowledge. In this study, we frame exclusion as a data-driven estimation problem, and apply flexible machine learning methods to estimate the probability of a unit complying with the instrument. We demonstrate how excluding likely noncompliers can increase power while maintaining valid treatment effect estimates. We show the utility of our approach with a fuzzy regression discontinuity analysis of the effect of initial diabetes diagnosis on follow-up blood sugar levels. Data-driven exclusion criterion can help improve both power and external validity for various quasi-experimental settings.

PDF Details

AAAI Conference 2019 Conference Paper

Unsupervised Post-Processing of Word Vectors via Conceptor Negation

Tianlin Liu
Lyle Ungar
João Sedoc

Word vectors are at the core of many natural language processing tasks. Recently, there has been interest in postprocessing word vectors to enrich their semantic information. In this paper, we introduce a novel word vector postprocessing technique based on matrix conceptors (Jaeger 2014), a family of regularized identity maps. More concretely, we propose to use conceptors to suppress those latent features of word vectors having high variances. The proposed method is purely unsupervised: it does not rely on any corpus or external linguistic database. We evaluate the post-processed word vectors on a battery of intrinsic lexical evaluation tasks, showing that the proposed method consistently outperforms existing state-of-the-art alternatives. We also show that post-processed word vectors can be used for the downstream natural language processing task of dialogue state tracking, yielding improved results in different dialogue domains.

PDF Details

AAAI Conference 2017 Conference Paper

Proper Proxy Scoring Rules

Jens Witkowski
Pavel Atanasov
Lyle Ungar
Andreas Krause

Proper scoring rules can be used to incentivize a forecaster to truthfully report her private beliefs about the probabilities of future events and to evaluate the relative accuracy of forecasters. While standard scoring rules can score forecasts only once the associated events have been resolved, many applications would beneﬁt from instant access to proper scores. In forecast aggregation, for example, it is known that using weighted averages, where more weight is put on more accurate forecasters, outperforms simple averaging of forecasts. We introduce proxy scoring rules, which generalize proper scoring rules and, given access to an appropriate proxy, allow for immediate scoring of probabilistic forecasts. In particular, we suggest a proxy-scoring generalization of the popular quadratic scoring rule, and characterize its incentive and accuracy evaluation properties theoretically. Moreover, we thoroughly evaluate it experimentally using data from a large real world geopolitical forecasting tournament, and show that it is competitive with proper scoring rules when the number of questions is small.

PDF Details

AAAI Conference 2016 Conference Paper

Discovering User Attribute Stylistic Differences via Paraphrasing

Daniel Preotiuc-Pietro
Wei Xu
Lyle Ungar

User attribute prediction from social media text has proven successful and useful for downstream tasks. In previous studies, differences in user trait language use have been limited primarily to the presence or absence of words that indicate topical preferences. In this study, we aim to ﬁnd linguistic style distinctions across three different user attributes: gender, age and occupational class. By combining paraphrases with a simple yet effective method, we capture a wide set of stylistic differences that are exempt from topic bias. We show their predictive power in user proﬁling, conformity with human perception and psycholinguistic hypotheses, and potential use in generating natural language tailored to speciﬁc user traits.

PDF Details

JMLR Journal 2014 Journal Article

Spectral Learning of Latent-Variable PCFGs: Algorithms and Sample Complexity

Shay B. Cohen
Karl Stratos
Michael Collins
Dean P. Foster
Lyle Ungar

We introduce a spectral learning algorithm for latent-variable PCFGs (Matsuzaki et al., 2005; Petrov et al., 2006). Under a separability (singular value) condition, we prove that the method provides statistically consistent parameter estimates. Our result rests on three theorems: the first gives a tensor form of the inside- outside algorithm for PCFGs; the second shows that the required tensors can be estimated directly from training examples where hidden-variable values are missing; the third gives a PAC-style convergence bound for the estimation method. [abs] [ pdf ][ bib ] &copy JMLR 2014. ( edit, beta )

PDF Details

NeurIPS Conference 2013 Conference Paper

Faster Ridge Regression via the Subsampled Randomized Hadamard Transform

Yichao Lu
Paramveer Dhillon
Dean Foster
Lyle Ungar

We propose a fast algorithm for ridge regression when the number of features is much larger than the number of observations ($p \gg n$). The standard way to solve ridge regression in this setting works in the dual space and gives a running time of $O(n^2p)$. Our algorithm (SRHT-DRR) runs in time $O(np\log(n))$ and works by preconditioning the design matrix by a Randomized Walsh-Hadamard Transform with a subsequent subsampling of features. We provide risk bounds for our SRHT-DRR algorithm in the fixed design setting and show experimental results on synthetic and real datasets.

PDF Details

NeurIPS Conference 2013 Conference Paper

New Subsampling Algorithms for Fast Least Squares Regression

Paramveer Dhillon
Yichao Lu
Dean Foster
Lyle Ungar

We address the problem of fast estimation of ordinary least squares (OLS) from large amounts of data ($n \gg p$). We propose three methods which solve the big data problem by subsampling the covariance matrix using either a single or two stage estimation. All three run in the order of size of input i. e. O($np$) and our best method, {\it Uluru}, gives an error bound of $O(\sqrt{p/n})$ which is independent of the amount of subsampling as long as it is above a threshold. We provide theoretical bounds for our algorithms in the fixed design (with Randomized Hadamard preconditioning) as well as sub-Gaussian random design setting. We also compare the performance of our methods on synthetic and real-world datasets and show that if observations are i. i. d. , sub-Gaussian then one can directly subsample without the expensive Randomized Hadamard preconditioning without loss of accuracy.

PDF Details

NeurIPS Conference 2011 Conference Paper

Multi-View Learning of Word Embeddings via CCA

Paramveer Dhillon
Dean Foster
Lyle Ungar

Recently, there has been substantial interest in using large amounts of unlabeled data to learn word representations which can then be used as features in supervised classifiers for NLP tasks. However, most current approaches are slow to train, do not model context of the word, and lack theoretical grounding. In this paper, we present a new learning method, Low Rank Multi-View Learning (LR-MVL) which uses a fast spectral method to estimate low dimensional context-specific word representations from unlabeled data. These representation features can then be used with any supervised learner. LR-MVL is extremely fast, gives guaranteed convergence to a global optimum, is theoretically elegant, and achieves state-of-the-art performance on named entity recognition (NER) and chunking problems.

PDF Details

YNIMG Journal 2010 Journal Article

Dementia induces correlated reductions in white matter integrity and cortical thickness: A multivariate neuroimaging study with sparse canonical correlation analysis

Brian B. Avants
Philip A. Cook
Lyle Ungar
James C. Gee
Murray Grossman

We use a new, unsupervised multivariate imaging and analysis strategy to identify related patterns of reduced white matter integrity, measured with the fractional anisotropy (FA) derived from diffusion tensor imaging (DTI), and decreases in cortical thickness, measured by high resolution T1-weighted imaging, in Alzheimer's disease (AD) and frontotemporal dementia (FTD). This process is based on a novel computational model derived from sparse canonical correlation analysis (SCCA) that allows us to automatically identify mutually predictive, distributed neuroanatomical regions from different imaging modalities. We apply the SCCA model to a dataset that includes 23 control subjects that are demographically matched to 49 subjects with autopsy or CSF-biomarker-diagnosed AD (n =24) and FTD (n =25) with both DTI and T1-weighted structural imaging. SCCA shows that the FTD-related frontal and temporal degeneration pattern is correlated across modalities with permutation corrected p <0. 0005. In AD, we find significant association between cortical thinning and reduction in white matter integrity within a distributed parietal and temporal network (p <0. 0005). Furthermore, we show that—within SCCA identified regions—significant differences exist between FTD and AD cortical-connective degeneration patterns. We validate these distinct, multimodal imaging patterns by showing unique relationships with cognitive measures in AD and FTD. We conclude that SCCA is a potentially valuable approach in image analysis that can be applied productively to distinguishing between neurodegenerative conditions.

Details DOI

NeurIPS Conference 2008 Conference Paper

Regularized Learning with Networks of Features

Ted Sandler
John Blitzer
Partha Talukdar
Lyle Ungar

For many supervised learning problems, we possess prior knowledge about which features yield similar information about the target variable. In predicting the topic of a document, we might know that two words are synonyms, or when performing image recognition, we know which pixels are adjacent. Such synonymous or neighboring features are near-duplicates and should therefore be expected to have similar weights in a good model. Here we present a framework for regularized learning in settings where one has prior knowledge about which features are expected to have similar and dissimilar weights. This prior knowledge is encoded as a graph whose vertices represent features and whose edges represent similarities and dissimilarities between them. During learning, each feature's weight is penalized by the amount it differs from the average weight of its neighbors. For text classification, regularization using graphs of word co-occurrences outperforms manifold learning and compares favorably to other recently proposed semi-supervised learning methods. For sentiment analysis, feature graphs constructed from declarative human knowledge, as well as from auxiliary task learning, significantly improve prediction accuracy.

PDF Details

NeurIPS Conference 2001 Conference Paper

Rates of Convergence of Performance Gradient Estimates Using Function Approximation and Bias in Reinforcement Learning

Gregory Grudic
Lyle Ungar

We address two open theoretical questions in Policy Gradient Reinforce- ment Learning. The ﬁrst concerns the efﬁcacy of using function approx- imation to represent the state action value function, . Theory is pre- sented showing that linear function approximation representations of can degrade the rate of convergence of performance gradient estimates by a factor of relative to when no function approximation of is used, where is the number of basis functions in the function approximation representation. The sec- ond concerns the use of a bias term in estimating the state action value function. Theory is presented showing that a non-zero bias term can improve the rate of convergence of performance gradient estimates by is the number of possible actions. Experimen- tal evidence is presented showing that these theoretical results lead to signiﬁcant improvement in the convergence properties of Policy Gradi- ent Reinforcement Learning algorithms. is the number of possible actions and

PDF Details