Publications
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Our teams aspire to make discoveries that impact everyone, and core to our approach is sharing our research and tools to fuel progress in the field.
Sort By
1 - 15 of 11371 publications
Preview abstract
Optimizing large-language model (LLM) training and serving on large-sacle distributed systems with hundreds and thousands of accelerators is always a challenging task due to the fast evloving LLMs, strong domain expertise required, and various optimization goals from different worklaods. Existing methods rely on either handcrafted optimization performed by human experts, which is tedious and time-consuming or resource-intensive black-box searches, which lack the extensibility to keep pace with evolving models and hardware. To address this, we introduce PROMPTS, a novel multi-agent framework that complements traditional search methods with expert-informed reasoning. It automates the diagnosis of performance bottlenecks by synthesizing profiler data and leverages a knowledge base to propose optimized sharding configurations with detailed justifications.
Across eight real-world production workloads, PROMPTS demonstrated remarkable efficiency and accuracy, delivering performance improvements of up to 434%. These workloads spanned diverse model architectures, hardware platforms, computational scales, and various stages of the machine learning lifecycle (pre-training, serving, and post-training). In every case, the configuration adopted by human engineers was identified within the agent's top three proposals from a single invocation. Furthermore, the agent's top-ranked recommendation was the one ultimately adopted in 87.5% of cases, showcasing its ability to not only find optimized solutions, but also to correctly prioritize them. Our work establishes PROMPTS as a scalable, extensible, and explainable methodology for AI-assisted performance engineering in large-scale ML systems.
View details
Preview abstract
Multimodal large language models (LLMs) integrate and process information from multiple modalities such as text, images, audio, and video, enabling complex tasks such as audio translation and visual question answering. While powerful, this complexity introduces novel vulnerabilities to sophisticated adversarial attacks. This survey paper provides a comprehensive overview of this rapidly expanding field, systematically categorizing attacks that range from manipulations of single modalities (e.g., perturbed images or audio) to those exploiting cross-modal interactions. We overview how these attacks exploit weaknesses in model fusion, attention mechanisms, and representation learning and provided analyses on their potential for real-world consequences.
View details
Digital Shadow AI Risk Theoretical Framework (DART): A Framework for Managing Data Disclosure and Privacy Risks of AI tools at Work
Master's Thesis (2026)
Preview abstract
The accelerated integration of generative AI technologies and agentic AI tools, particularly those like ChatGPT, into workplace settings has introduced complex challenges concerning data governance, regulatory compliance, and organizational privacy (GDPR 2016; CCPA/CPRA). This study introduces the Digital Shadow AI Risk Theoretical Framework (DART)—a novel theoretical framework designed to systematically identify, classify, and address the latent risks arising from the widespread, and often unregulated, use of AI systems in professional environments (NIST, 2023; OECD AI Policy Observatory, 2023). DART introduces six original, interrelated constructs developed in this study: Unintentional Disclosure Risk, Trust-Dependence Paradox, Data Sovereignty Conflict, Knowledge Dilution Phenomenon, Ethical Black Box Problem, and Organizational Feedback Loops. Each construct reflects a unique dimension of risk that emerges as organizations increasingly rely on AI-driven tools for knowledge work and decision-making.
The framework is empirically tested through a mixed-methods research design involving hypothesis testing and statistical analysis of behavioral data gathered from cross-sectional surveys of industry professionals. Two cross-industry surveys (Survey-1: 416 responses, 374 analyzed; Survey-2: 203 responses, 179 analyzed) and CB-SEM tests supported seven of eight hypotheses; H4 (sovereignty) was not significant; H7 (knowledge dilution) was confirmed in replication. The findings highlight critical gaps in employee training, policy awareness, and risk mitigation strategies—underscoring the urgent need for updated governance frameworks, comprehensive AI-use policies, and targeted educational interventions. This paper contributes to emerging scholarship by offering a robust model for understanding and mitigating digital risks in AI-enabled workplaces, providing practical implications for compliance officers, risk managers, and organizational leaders aiming to harness the benefits of generative AI responsibly and securely. The novelty of DART lies in its explicit theorization of workplace-level behavioral risks—especially Shadow AI, which unlike Shadow IT externalizes organizational knowledge into adaptive systems—thereby offering a unified framework that bridges fragmented literatures and grounds them in empirical evidence.
View details
MoXaRt: Audio-Visual Object-Guided Sound Interaction for XR
Sieun Kim
Qianhui Zheng
Ruoyu Xu
Ravi Tejasvi
Anuva Kulkarni
Junyi Zhu
2026
Preview abstract
In Extended Reality (XR), complex acoustic environments often overwhelm users, compromising both scene awareness and social engagement due to entangled sound sources. We introduce MoXaRt, a real-time XR system that uses audio-visual cues to separate these sources and enable fine-grained sound interaction. MoXaRt's core is a cascaded architecture that performs coarse, audio-only separation in parallel with visual detection of sources (e.g. faces, instruments). These visual anchors then guide refinement networks to isolate individual sources, separating complex mixes of up to five concurrent sources (e.g. two voices + three instruments) with ca. 2 second processing latency. We validate MoXaRt through a technical evaluation on a new, complex dataset we collected, and a 22-participant user study. Our results demonstrate that MoXaRt significantly improves communication clarity—boosting listening comprehension in noisy conditions by 33.2% (p=0.0058)—and significantly reduces cognitive load (M=7.50 vs. M=3.36, p<0.001), paving the way for more perceptive and socially adept XR experiences.
View details
Preview abstract
Source-to-source compilers may perform inefficiently by executing transpilation passes on scripts that do not contain the specific language features a pass is designed to transform, potentially leading to redundant processing. A compiler can analyze a script to generate a per-script feature map, for example, by identifying language features in its abstract syntax tree (AST). Before executing a transpilation pass, the compiler can check this map and may bypass the pass for that script if the specific feature targeted by the pass is not present. This feature map can also be dynamically updated throughout the compilation process as other passes transform the code. This method of conditional pass execution based on content-aware analysis may reduce redundant AST traversals, which could decrease overall compilation time and computational resource consumption.
View details
Preview abstract
In some multi-stage software build pipelines, downstream compiler errors may be reported against ephemeral, machine-generated intermediate artifacts rather than original, human-written source code, which can make remediation challenging. A system and method may address this by intercepting a downstream error, mapping its location back to the original source file, and programmatically injecting a dormant suppression tag into the original source code. During a subsequent build, an intermediate transpiler can propagate this tag into a newly generated intermediate artifact. In the intermediate file, the tag may become active and be recognized by the downstream compiler as a directive to suppress the specific error. This approach can facilitate an automated remediation process for certain build failures that avoids direct modification of ephemeral files and uses the original source code as a record for suppression.
View details
Preview abstract
How many T gates are needed to approximate an arbitrary n-qubit quantum state to within
a given precision ϵ? Improving prior work of Low, Kliuchnikov and Schaeffer, we show that the
optimal asymptotic scaling is Θ(sqrt{2^n log(1/ε)} + log(1/ε)) if we allow an unlimited number of ancilla qubits. We also show that this is the optimal T-count for implementing an arbitrary
diagonal n-qubit unitary to within error ϵ. We describe an application to batched synthesis of
single-qubit unitaries: we can approximate a tensor product of m = O(log log(1/ϵ)) arbitrary
single-qubit unitaries to within error ϵ with the same asymptotic T-count as is required to
approximate just one single-qubit unitary.
View details
Towards AI as a Collaborative Partner: A Taxonomy of AI Agent Behavior in Software Engineering
Proceedings of the 3rd ACM International Conference on AI-Powered Software (AIware '26), ACM, Montreal, QC, Canada (2026) (to appear)
Preview abstract
The ongoing transition of Large Language Models (LLMs) in software engineering from one-shot code generators into agentic partners requires a shift in how we define and measure success. While models are becoming more capable, the industry lacks a clear understanding of the behavioral norms that make an interactive software engineering (SWE) agent effective in collaborative software development in the enterprise. This work addresses this gap by presenting a taxonomy of desirable SWE agent behaviors, synthesized from 91 sets of developer-defined rules for SWE agents and validated through interviewing 15 experienced professional developers. In this taxonomy, we identify four core expectations: Adhere to Standards and Processes, Ensure Code Quality and Reliability, Solve Problems Effectively, and Collaborate with the Developer. These findings offer a concrete vocabulary for aligning SWE agent behavior with developer preferences, enabling researchers and practitioners to move beyond correctness-only benchmarks and start designing evaluations that reflect the socio-technical nature of professional software development in enterprises.
View details
Preview abstract
Online video platforms face an exponential challenge in detecting and mitigating the flood of AI-generated "slop" and synthetic spam perpetuated by coordinated malicious actors. This content is increasingly designed to exploit the limitations of traditional media forensics, often utilizing generative AI to produce unique, localized variations of harmful or low-quality material at scale. Traditional content-centric moderation fails against this coordinated, adversarial generation strategy.
This paper presents a novel, scalable defense system deployed at a major Online Video Platform (OVP) to identify and terminate clusters of coordinated accounts exhibiting a prevalence of adversarial synthetic content. The approach leverages a multi-faceted architecture incorporating two core machine learning components: a robust Coordinated Bot-Net Detector (via Account Relatedness) and a Synthetic Pattern Classifier (formerly BT Classifier). Crucially, we introduce an advanced AI enhancement layer utilizing Large Language Models (LLMs), specialized via Low-Rank Adaptation (LoRA) and Automatic Prompt Optimization (APO), to achieve rapid, high-precision semantic understanding of emerging synthetic spam trends.
Operational data spanning a six-month period demonstrates the system's significant impact, resulting in the successful termination of 50K clusters comprising 130K channels of synthetic spam generators. Furthermore, the LLM-driven automation significantly improves operational efficiency, saving approximately 83 human review hours to cut down human reviews by 50%. This work details a critical, deployed solution that provides essential scalability and adversarial resilience against sophisticated generative attacks.
View details
ConvApparel: A Benchmark Dataset and Validation Framework for User Simulators in Conversational Recommenders
Ofer Meshi
Guy Tennenholtz
Jihwan Jeong
Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL-26), Rabat, Morocco (2026), pp. 5270-5304
Preview abstract
LLM-based user simulators are a scalable solution for improving conversational AI, but a critical realism gap undermines their effectiveness. To close this gap, we introduce a framework for building and validating high-fidelity simulators. We present a novel dataset of human-AI shopping conversations designed to capture a wide spectrum of user experiences. To measure fidelity, we propose a hybrid evaluation protocol that combines statistical alignment with a learned, discriminator-based Human-Likeness Score. Our most sophisticated simulator, trained via reinforcement learning with iterative critique, achieves a significant leap in realism. Critically, we demonstrate through counterfactual validation that our simulator—trained exclusively on optimal interactions—realistically adapts its behavior to suboptimal system responses, mirroring real user reactions and marking a key advance in creating reliable simulators for robust AI development.
View details
Preview abstract
We introduce a new context-enriched time series forecasting benchmark TimesX. TimesX contains a wide selection of high-quality real-world time series and diverse textual contexts from an automated generating pipeline, which helps address three main issues of existing benchmarks: (1) poor generalization due to low data volume and data being synthetic, (2) restricted forms of context, and (3) an inability to mitigate data leakage. We conduct a thorough empirical study of current multimodal solutions on TimesX. Our results suggest that most multimodal solutions that work well on existing benchmarks may fail on TimesX. In contrast, simple ensemble methods that leverage the rich textual context can outperform strong unimodal baselines and other multimodal baselines.
** Below this is what was submitted to ITP. **
We create a real world multimodal time-series forecasting benchmark that encompasses diverse domains and regions. Each time-series is annotated by various kinds of contexts like metadata, date and holiday information, dynamic events related to the time-series. This is sufficiently more advanced than other available benchmarks which rely wither on static metadata alone or synthetic examples. This forms a test bed for multimodal forecasting. We also present some baseline results showing that ensembles of publicly available LLMs and time-series foundation models can demonstrate non-trivial performance on this bechmark.
View details
Preview abstract
PURPOSE: To introduce Cardio Load (CL), a metric quantifying cardiovascular work from all activities across the day, and to investigate its distribution by age, gender, and workout profiles. CL adapts the Training Impulse (TRIMP) model by leveraging continuous heart rate and movement data from wearables, enabling minute-level intensity estimation. We also discuss the derivation of weekly target loads, intended to guide fitness maintenance.
METHODS: A retrospective analysis was conducted on 31.2 million hours of wrist-worn wearable data collected over a six-week period. The dataset comprised a 40,000-subject subset (37.9% female) of consenting Google Pixel Watch® users in the United States, aged 18 to 80 years (18-39: 41.8%, 40-59: 43.5%, 60+: 14.6%). Measured data included minute-interval heart rate averages, resting and maximum heart rates, minute-interval averaged accelerometer log energy, and manually-logged or auto-detected activity types. Cardio Load scores and target loads were calculated daily for each subject and compared across age and gender. We also compared the proportions of CL gained during workouts and incidental daily activities for these groups.
RESULTS: Overall, the study population's mean ± SD weekly CL scores were 221 ± 156 (female) and 259 ± 169 (male). Median weekly Cardio Load (CL) values exhibited consistency for individuals between 30 and 75 years of age. When analyzed in five-year age groups, the coefficient of variation (CV%) of median weekly CL values within this age range was less than 4.5%, with younger and older subjects demonstrating higher and lower median CL, respectively. The median proportion of CL accumulated during structured workouts versus incidental daily activity was 41.0% (female) and 49.0% (male) for all subjects, though this varied considerably with average weekly workout duration. CV% of weekly target load and daily target load over 6 weeks was 23.6% and 35.2% respectively.
CONCLUSION: Cardio Load provides a continuous quantification of activity load from wearables, acknowledging both structured workouts and everydayincidental activity. CL is equitably rewarded for age ranges spanning 30-75 years. Weekly target loads were found to have little measurement variability and be more consistent and, consequently, more practical for planning training and physical activity than daily targets.
View details
Preview abstract
Voice activity detection (VAD) plays a vital role in enabling applications such as speech recognition. We analyze the impact of window size on the accuracy of three VAD algorithms: Silero, WebRTC, and Root Mean Square (RMS) across a set of diverse real-world digital audio streams. We additionally explore the use of hysteresis on top of each VAD output. Our results offer practical references for optimizing VAD systems. Silero significantly outperforms WebRTC and RMS, and hysteresis provides a benefit for WebRTC.
View details
Preview abstract
Source-to-source compilers may perform inefficiently by executing transpilation passes on scripts that do not contain the specific language features a pass is designed to transform, potentially leading to redundant processing. A compiler can analyze a script to generate a per-script feature map, for example, by identifying language features in its abstract syntax tree (AST). Before executing a transpilation pass, the compiler can check this map and may bypass the pass for that script if the specific feature targeted by the pass is not present. This feature map can also be dynamically updated throughout the compilation process as other passes transform the code. This method of conditional pass execution based on content-aware analysis may reduce redundant AST traversals, which could decrease overall compilation time and computational resource consumption.
View details
In Search of "Weird Corners": Diagnosing the Limits of Convergent AI in Professional Creative Practice
Mark Williams
2026
Preview abstract
Generative AI assistants typically employ convergent interaction paradigms to resolve ambiguity. While effective for technical tasks, this risks premature convergence in creative domains, constraining output variance. Evaluating a convergent AI probe with expert creatives (N=9) indicates an interactional paradox: structural linearity provides "ignition" utility for early ideation, but misaligns with organic workflows, often inducing "aesthetic sanitization" that standardizes individualized nuance. Prioritizing constructive friction over default agreement, the experts requested active, lateral collaborators. In response, we reframe output convergence as a "full-stack" UI challenge, advocating for Generative frameworks that operationalize the Double Diamond via fluid role-shifting and productive tension.
View details
