Dennis Wei

Dennis Wei

Auto-generated profile, based on publicly available information

Dennis Wei

IBM Research The Chinese University of Hong Kong Harvard University

Research Focus & Metrics

Academic Career Length

13Years (2014-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 January 2025
Verified
Conference Paper
Open access
PDF file
LPLucas Monteiro Paes
Lucas Monteiro Paes
DWDennis Wei
Dennis Wei
HDHyo Jin Do
Hyo Jin Do
HSHendrik Strobelt
Hendrik Strobelt
RLRonny Luss
Ronny Luss
SGSoumya Sankar Ghosh
Soumya Sankar Ghosh
Lucas Monteiro Paes, Dennis Wei, Hyo Jin Do, Hendrik Strobelt, Ronny Luss, Amit Dhurandhar, Manish Nagireddy, Karthikeyan Natesan Ramamurthy, Prasanna Sattigeri, Werner Geyer, Soumya Ghosh. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

11 April 2017 in arXiv
Verified
Preprint
Open access
PDF file
FCFlavio P. Calmon
Flavio P. Calmon
DWDennis Wei
Dennis Wei
KRKarthikeyan Natesan Ramamurthy
Karthikeyan Natesan Ramamurthy
KVKush Varshney
Kush Varshney
Non-discrimination is a recognized objective in algorithmic decision making. In this paper, we introduce a novel probabilistic formulation of data pre-processing for reducing discrimination. We propose a convex optimization for learning a data transformation with three goals: controlling discrimination, limiting distortion in individual data samples, and preserving utility. We characterize the impact of limited sample size in accomplishing this objective, and apply two instances of the proposed optimization to datasets, including one on real-world criminal recidivism. The results demonstrate that all three criteria can be simultaneously achieved and also reveal interesting patterns of bias in American society.

6 August 2020 in arXiv
Verified
Preprint
Open access
PDF file
SDSanjeeb Dash
Sanjeeb Dash
OGOktay Günlük
Oktay Günlük
DWDennis Wei
Dennis Wei
This paper considers the learning of Boolean rules in either disjunctive normal form (DNF, OR-of-ANDs, equivalent to decision rule sets) or conjunctive normal form (CNF, AND-of-ORs) as an interpretable model for classification. An integer program is formulated to optimally trade classification accuracy for rule simplicity. Column generation (CG) is used to efficiently search over an exponential number of candidate clauses (conjunctions or disjunctions) without the need for heuristic rule mining. This approach also bounds the gap between the selected rule set and the best possible rule set on the training data. To handle large datasets, we propose an approximate CG algorithm using randomization. Compared to three recently proposed alternatives, the CG algorithm dominates the accuracy-simplicity trade-off in 7 out of 15 datasets. When maximized for accuracy, CG is competitive with rule learners designed for this purpose, sometimes finding significantly simpler solutions that are no less accurate.

15 October 2025 in Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society
Verified
Journal Article
Open access
PDF file
HDHyo Jin Do
Hyo Jin Do
RORachel Ostrand
Rachel Ostrand
WGWerner Geyer
Werner Geyer
KMKeerthiram Murugesan
Keerthiram Murugesan
DWDennis Wei
Dennis Wei
JWJustin Weisz
Justin Weisz
Large language models (LLMs) are susceptible to generating inaccurate or false information, often referred to as "hallucinations" or "confabulations." While several technical advancements have been made to detect hallucinated content by assessing the factuality of the model's responses, there is still limited research on how to effectively communicate this information to users. To address this gap, we conducted two scenario-based experiments with a total of 208 participants to systematically compare the effects of various design strategies for communicating factuality scores by assessing participants' ratings of trust, ease in validating response accuracy, and preference. Our findings reveal that participants preferred and trusted a design in which all phrases within a response were color-coded based on factuality scores. Participants also found it easier to validate accuracy of the response in this style compared to a baseline with no style applied. Our study offers practical design guidelines for LLM application developers and designers, aimed at calibrating user trust, aligning with user preferences, and enhancing users' ability to scrutinize LLM outputs.

9 August 2025 in arXiv
Verified
Preprint
Open access
PDF file
HDHyo Jin Do
Hyo Jin Do
RORachel Ostrand
Rachel Ostrand
WGWerner Geyer
Werner Geyer
KMKeerthiram Murugesan
Keerthiram Murugesan
DWDennis Wei
Dennis Wei
JWJustin Weisz
Justin Weisz
Large language models (LLMs) are susceptible to generating inaccurate or false information, often referred to as "hallucinations" or "confabulations." While several technical advancements have been made to detect hallucinated content by assessing the factuality of the model's responses, there is still limited research on how to effectively communicate this information to users. To address this gap, we conducted two scenario-based experiments with a total of 208 participants to systematically compare the effects of various design strategies for communicating factuality scores by assessing participants' ratings of trust, ease in validating response accuracy, and preference. Our findings reveal that participants preferred and trusted a design in which all phrases within a response were color-coded based on factuality scores. Participants also found it easier to validate accuracy of the response in this style compared to a baseline with no style applied. Our study offers practical design guidelines for LLM application developers and designers, aimed at calibrating user trust, aligning with user preferences, and enhancing users' ability to scrutinize LLM outputs.

4 August 2014 in arXiv
Verified
Preprint
Open access
PDF file
DWDennis Wei
Dennis Wei
AHAlfred O Hero
Alfred O Hero
This paper studies adaptive sensing for estimating the nonzero amplitudes of a sparse signal with the aim of providing analytical guarantees on the performance gain due to adaptive resource allocation. We consider a previously proposed optimal two-stage policy for allocating sensing resources. For positive powers q, we derive tight upper bounds on the mean qth-power error resulting from the optimal two-stage policy and corresponding lower bounds on the improvement over non-adaptive uniform sensing. It is shown that the adaptation gain is related to the detectability of nonzero signal components as characterized by Chernoff coefficients, thus quantifying analytically the dependence on the sparsity level of the signal, the signal-to-noise ratio, and the sensing resource budget. For fixed sparsity levels and increasing signal-to-noise ratio or sensing budget, we obtain the rate of convergence to oracle performance and the rate at which the fraction of resources spent on the first exploratory stage decreases to zero. For a vanishing fraction of nonzero components, the gain increases without bound as a function of signal-to-noise ratio and sensing budget. Numerical simulations demonstrate that the bounds on adaptation gain are quite tight in non-asymptotic regimes as well.