About

Tianyi Chen is an associate professor of electrical and computer engineering at Cornell Tech and Cornell Engineering. Prior to joining Cornell Tech, he was an associate professor at Rensselaer Polytechnic Institute (RPI), where his research was jointly supported by the RPI–IBM Artificial Intelligence Research Partnership. Chen received his bachelor’s degree from Fudan University in 2014 and his doctoral degree from the University of Minnesota in 2019. His research focuses on the theoretical foundations and algorithmic design of multi-objective optimization and learning, with applications to generative AI and emerging computing paradigms such as distributed and analog computing. His work bridges theory and real-world impact, leading to several joint patents with IBM. Chen is the inaugural recipient of the IEEE Signal Processing Society Best Ph.D. Dissertation Award in 2020 and a recipient of the NSF CAREER Award in 2021. His research has also been recognized with multiple industrial awards, including the Amazon Research Award and the Cisco Research Award. He has received several best paper recognitions, including the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) Best Student Paper Award in 2021 and the IEEE Signal Processing Society Young Author Best Paper Award in 2024.

Research topics

• Computer Science
• Artificial Intelligence
• Machine Learning
• Mathematics
• Computer hardware
• Data science
• Psychology
• Multimedia
• Programming language
• Computer graphics (images)
• Medicine
• Communication
• Medical physics

Selected publications

• Association between rheumatoid arthritis and periodontitis: a study based on a two-sample mendelian randomisation analysis

  Medicina oral, patología oral y cirugía bucal · 2025-01-01 · 2 citations

  articleOpen access

  BACKGROUND: The association between Rheumatoid arthritis (RA) and Periodontitis (PD) has been increasingly recognised, yet traditional epidemiological studies face challenges in establishing associations. Therefore, this study aims to genetically assess the association between RA and PD through Mendelian randomisation (MR) analysis, using genetic variations as instrumental variables. MATERIAL AND METHODS: Data on RA and PD were downloaded from the EBI website. The RA data contained 8,255 cases and 409,001 controls, with a total of 24,175,266 SNPs; the chronic PD data contained 950 cases and 409,001 controls, with a total of 11,842,647 SNPs; the acute PD data contained 128 cases and 456,220 controls, with a total of 11,842,647 SNPs. Additionally, the potential association between RA and PD was investigated. The intercept between Mendelian randomisation (MR)-Egger regression, MR-PRESSO test results and funnel plots was used to analyse the horizontal pleiotropy of SNPs along with the effect of individual SNPs on inverse-variance weighting (IVW) analysis results, assessed using the leave-one-out method. RESULTS: In total, 26 SNPs highly associated with RA were screened; MR-Egger regression (OR=1.242, 95% CI (1.032-1.494), P=0.031), WM (OR=1.190, 95% CI (1.015-1.395), P=0.032), IVW (OR=1.191, 95% CI (1.053-1.348), P=0.006) and weighted mode (OR=1.212, 95% CI (1.043-1.409), P=0.019) suggested that RA was a likelihood factor for chronic PD, whereas RA was not associated with the incidence of acute PD, and the Cochran's Q test indicated no statistical heterogeneity between SNPs highly associated with RA. Moreover, analyses using the intercept between the MR-Egger regression, MR-PRESSO test results and funnel plot revealed no horizontal pleiotropy in SNPs highly associated with RA. CONCLUSIONS: Rheumatoid arthritis was genetically identified as a likelihood factor for PD and the onset of chronic PD, but no association was observed between RA and acute PD.

  PublisherDOI

• FSL-Rectifier: Rectify Outliers in Few-Shot Learning via Test-Time Augmentation

  Proceedings of the AAAI Conference on Artificial Intelligence · 2025-04-11

  articleOpen accessSenior author

  Few-shot learning (FSL) commonly requires a model to identify images (queries) that belong to classes unseen during training, based on a few labelled samples of the new classes (support set) as reference. So far, plenty of algorithms involve training data augmentation to improve the generalization capability of FSL models, but outlier queries or support images during inference can still pose great generalization challenges. In this work, to reduce the bias caused by the outlier samples, we generate additional test-class samples by combining original samples with suitable train-class samples via a generative image combiner. Then, we obtain averaged features via an augmentor, which leads to more typical representations through the averaging. We experimentally and theoretically demonstrate the effectiveness of our method, obtaining a test accuracy improvement proportion of around 10% (e.g., from 46.86% to 53.28%) for trained FSL models. Importantly, given a pretrained image combiner, our method is training-free for off-the-shelf FSL models, whose performance can be improved without extra datasets nor further training of the models themselves.

  PublisherOA PDFDOI

• DSAC-C: Robust Discrete Soft-Actor Critic for Distribution Shift Scenarios

  2025-06-30

  articleSenior author

  We present a novel extension to the family of Soft Actor-Critic (SAC) algorithms. We argue that based on the Maximum Entropy Principle, discrete SAC can be further improved via additional statistical constraints derived from a surrogate critic policy. Furthermore, by perturbing these constraints we demonstrate that an added robustness can be provided against potentially unseen or adversarial environments, which are essential for safe deployment of reinforcement learning agents in the real-world. We provide theoretical analysis and show empirical results on low data regimes for both in-distribution and out-of-distribution variants of Atari 2600 games.

  PublisherDOI

• Monitoring for classical swine fever virus persistence and seropositivity in vaccinated pig farms using on-farm sentinel pigs during the pre-elimination phase toward a CSF-free status

  Preventive Veterinary Medicine · 2025-07-01 · 1 citations

  articleOpen access

  Classical swine fever (CSF) is a transboundary viral disease that causes high mortality and systemic hemorrhages in domestic pigs and wild boars. In Taiwan, long-term implementation of compulsory vaccination has effectively prevented CSF, with no confirmed cases reported since 2006. On the path toward CSF elimination, a national surveillance program was initiated in 2021 for a two-year period. This program included monitoring unvaccinated on-farm sentinel pigs (with maternally derived antibody but without further vaccination with age) raised on 300 vaccinated farms annually, as well as dead pigs from rendering plants and culled sows from slaughterhouses. Sentinel pigs were sampled twice, at 3–12 weeks and 18–24 weeks of age, and tested for CSFV RNA and neutralizing antibodies (NA). CSFV vaccine strain RNA was detected in eight sentinel pigs from three farms, while all others were CSFV negative. The seropositive rates for anti-CSFV NAs in the first inspection were 68.0% in 2021 and 70.8% in 2022, and in the second inspection, they declined significantly to 10.8% and 9.5%, respectively. Seropositivity declined with age, indicating MDA waning without new infections. Among 1,277 dead pigs and 1,199 culled sows tested, tests for wild-type CSFV RNA were negative, while vaccine strain RNA was detected in 1.6 and 2.8% of dead pigs in 2021 and 2022, respectively. These results suggest that wild-type CSFV persistence is minimal under the current vaccination strategy. Overall, these results demonstrate the significance of the surveillance program, and Taiwan is well-positioned toward CSF elimination. • A CSF surveillance model was established using on-farm sentinel pigs. • All CSFV RNA-positive samples were confirmed as the LPC vaccine strain. • Seropositivity declined with age, indicating MDA waning without new infections. • High vaccine coverage in sows and pigs ensured strong herd immunity. • Monitoring dead pigs and culled sows aided in detecting persistent CSFV.

  PublisherDOI

• Abstract 118: Ischemic Stroke in a 47‐Year‐Old Male with Carotid Webs and Buschke‐Ollendorff syndrome

  Stroke Vascular and Interventional Neurology · 2024-11-01

  articleOpen access

  Carotid web (CW) is an atypical form of focal fibromuscular dysplasia defined as an abnormal‐shelf‐like projection of intimal fibrous tissue into the carotid bulb, and has the tendency to alter the hemodynamic flow, increasing the risk of thromboembolism, leading to recurrent strokes and/or TIAs. CW has been associated with a variety of different pathologies, including connective tissue disease, however, there are no reported links between Buschke‐Ollendorff syndrome (BOS) and CW. Here we report a case of stroke in a 47‐year‐old man with BOS, presenting with an R‐LVO with abnormal neovascularization and carotid webs on the left ICA, as visualized on cerebral angiogram.

  PublisherDOI

• MaxEnt Loss: Constrained Maximum Entropy for Calibration under Out-of-Distribution Shift

  Proceedings of the AAAI Conference on Artificial Intelligence · 2024-03-24 · 3 citations

  articleOpen accessSenior author

  We present a new loss function that addresses the out-of-distribution (OOD) network calibration problem. While many objective functions have been proposed to effectively calibrate models in-distribution, our findings show that they do not always fare well OOD. Based on the Principle of Maximum Entropy, we incorporate helpful statistical constraints observed during training, delivering better model calibration without sacrificing accuracy. We provide theoretical analysis and show empirically that our method works well in practice, achieving state-of-the-art calibration on both synthetic and real-world benchmarks. Our code is available at https://github.com/dexterdley/MaxEnt-Loss.

  PublisherOA PDFDOI

• FSL-QuickBoost: Minimal-Cost Ensemble for Few-Shot Learning

  2024-10-26

  articleOpen accessSenior author

  Few-shot learning (FSL) usually trains models on data from one set of classes, but tests them on data from a different set of classes, providing a few labeled support samples of the unseen classes as a reference for the trained model. Due to the lack of target-relevant training data, there is usually high generalization error with respect to the test classes. In this work, we conduct empirical explorations and propose an ensemble method (namely QuickBoost), which is efficient and effective for improving the generalization of FSL. Specifically, QuickBoost includes an alternative-architecture pretrained encoder with a one-vs-all binary classifier (namely FSL-Forest) based on random forest algorithm, and is ensembled with the off-the-shelf FSL models via logit-level averaging. Experiments on three benchmarks demonstrate that our method achieves state-of-the-art performance with good efficiency. Codes are available at https://github.com/WendyBaiYunwei/FSL-QuickBoost.

  PublisherOA PDFDOI

• VORD: Visual Ordinal Calibration for Mitigating Object Hallucinations in Large Vision-Language Models

  arXiv (Cornell University) · 2024-12-20

  preprintOpen accessSenior author

  Large Vision-Language Models (LVLMs) have made remarkable developments along with the recent surge of large language models. Despite their advancements, LVLMs have a tendency to generate plausible yet inaccurate or inconsistent information based on the provided source content. This phenomenon, also known as ``hallucinations" can have serious downstream implications during the deployment of LVLMs. To address this, we present VORD a simple and effective method that alleviates hallucinations by calibrating token predictions based on ordinal relationships between modified image pairs. VORD is presented in two forms: 1.) a minimalist training-free variant which eliminates implausible tokens from modified image pairs, and 2.) a trainable objective function that penalizes unlikely tokens. Our experiments demonstrate that VORD delivers better calibration and effectively mitigates object hallucinations on a wide-range of LVLM benchmarks.

  PublisherOA PDFDOI

• SFENet: Arbitrary Shapes Scene Text Detection with Semantic Feature Extractor

  Lecture notes in computer science · 2024-11-02

  book-chapter
  PublisherDOI

• Misinformation, Disinformation, and Generative AI: Implications for Perception and Policy

  Digital Government Research and Practice · 2024-08-23 · 20 citations

  articleOpen access

  The emergence of generative artificial intelligence (GenAI) has exacerbated the challenges of misinformation, disinformation, and mal-information (MDM) within digital ecosystems. These multi-faceted challenges demand a re-evaluation of the digital information lifecycle and a deep understanding of its social impact. An interdisciplinary strategy integrating insights from technology, social sciences, and policy analysis is crucial to address these issues effectively. This article introduces a three-tiered framework to scrutinize the lifecycle of GenAI-driven content from creation to consumption, emphasizing the consumer perspective. We examine the dynamics of consumer behavior that drive interactions with MDM, pinpoints vulnerabilities in the information dissemination process, and advocates for adaptive, evidence-based policies. Our interdisciplinary methodology aims to bolster information integrity and fortify public trust, equipping digital societies to manage the complexities of GenAI and proactively address the evolving challenges of digital misinformation. We conclude by discussing how GenAI can be leveraged to combat MDM, thereby creating a reflective cycle of technological advancement and mitigation.

  PublisherOA PDFDOI

Frequent coauthors

• Cha Zhang

  36 shared

• Andrew Gallagher

  Google (United States)

  36 shared

• Devi Parikh

  27 shared

• Adarsh Kowdle

  Google (United States)

  23 shared

• Simon Lucey

  21 shared

• Toshihiko Yamasaki

  20 shared

• Dhruv Batra

  17 shared

• Yao‐Jen Chang

  15 shared

Awards & honors

• IEEE Signal Processing Society Best Ph.D. Dissertation Award…
• NSF CAREER Award (2021)
• Amazon Research Award
• Cisco Research Award
• IEEE International Conference on Acoustics, Speech and Signa…