arXiv subjects

Tung Mai

Publications and source records attributed to Tung Mai.

At least 19 recordsLinked to original sources

arXiv2026

Constant Swap Regret in General-Sum Games via Optimistic Transition Matrices

We give deterministic and uncoupled learning dynamics for finite multiplayer general-sum games under full-information feedback that achieve constant individual swap regret, independent of the horizon $T$. With $n$ players and at most $m$ actions each, the individual swap regret of every player is $O(\sqrt{n} m \log m \log^{5/2}(nm))$ at every finite horizon. Each player predicts the deviation gains, then uses these predictions to update a row-stochastic transition matrix, and plays its stationary distribution. The proof combines a potential argument exploiting stationarity with a two-scale higher-order prediction analysis, using rooted-tree representations to handle the nonlinear dependence of deviation gains on the stationary distributions. An adversarially robust variant, obtained through a generic common-prefix switching wrapper, preserves the self-play bound up to a universal constant and guarantees individual swap regret at most $7\sqrt{m T \log m}$ in the adversarial setting.

cs.GT↗

arXiv2026

ML-Tool-Bench: Tool-Augmented Planning for ML Tasks

The development of autonomous machine learning (ML) agents capable of end-to-end data science workflows represents a significant frontier in artificial intelligence. These agents must orchestrate complex sequences of data analysis, feature engineering, model selection, and hyperparameter optimization, tasks that require sophisticated planning and iteration. While recent work on building ML agents has explored using large language models (LLMs) for direct code generation, tool-augmented approaches offer greater modularity and reliability. However, existing tool-use benchmarks focus primarily on task-specific tool selection or argument extraction for tool invocation, failing to evaluate the sophisticated planning capabilities required for ML Agents. In this work, we introduce a comprehensive benchmark for evaluating tool-augmented ML agents using a curated set of 61 specialized tools and 15 tabular ML challenges from Kaggle. Our benchmark goes beyond traditional tool-use evaluation by incorporating an in-memory named object management, allowing agents to flexibly name, save, and retrieve intermediate results throughout the workflows. We demonstrate that standard ReAct-style approaches struggle to generate valid tool sequences for complex ML pipelines, and that tree search methods with LLM-based evaluation underperform due to inconsistent state scoring. To address these limitations, we propose two simple approaches: 1) using shaped deterministic rewards with structured textual feedback, and 2) decomposing the original problem into a sequence of sub-tasks, which significantly improves trajectory validity and task performance. Using GPT-4o, our approach improves over ReAct by 16.52 percentile positions, taking the median across all Kaggle challenges. We believe our work provides a foundation for developing more capable tool-augmented planning ML agents.

cs.LG↗

arXiv2026

Robust Stable Matchings: Dealing with Changes in Preferences

We study stable matchings that are robust to preference changes in the two-sided stable matching setting of Gale and Shapley [GS62]. Given two instances $A$ and $B$ on the same set of agents, a matching is said to be robust if it is stable under both instances. This notion captures desirable robustness properties in matching markets where preferences may evolve, be misreported, or be subject to uncertainty. While the classical theory of stable matchings reveals rich lattice, algorithmic, and polyhedral structure for a single instance, it is unclear which of these properties persist when stability is required across multiple instances. Our work initiates a systematic study of the structural and computational behavior of robust stable matchings under increasingly general models of preference changes. We analyze robustness under a hierarchy of perturbation models: 1. a single upward shift in one agent's preference list, 2. an arbitrary permutation change by a single agent, and 3. arbitrary preference changes by multiple agents on both sides. For each regime, we characterize when: 1. the set of robust stable matchings forms a sublattice, 2. the lattice of robust stable matchings admits a succinct Birkhoff partial order enabling efficient enumeration, 3. worker-optimal and firm-optimal robust stable matchings can be computed efficiently, and 4. the robust stable matching polytope is integral (by studying its LP formulation). We provide explicit counterexamples demonstrating where these structural and geometric properties break down, and complement these results with XP-time algorithms running in $O(n^k)$ time, parameterized by $k$, the number of agents whose preferences change. Our results precisely delineate the boundary between tractable and intractable cases for robust stable matchings.

cs.GT↗

arXiv2025

Stable Matching: Dealing with Changes in Preferences

We study stable matchings that are robust to preference changes in the two-sided stable matching setting of Gale and Shapley[GS62]. Given two instances $A$ and $B$ on the same set of agents, a matching is said to be robust if it is stable under both instances. While prior work has considered the case where a single agent changes preferences between $A$ and $B$, we allow multiple agents on both sides to update their preferences and ask whether three central properties of stable matchings extend to robust stable matchings: (i) Can a robust stable matching be found in polynomial time? (ii) Does the set of robust stable matchings form a lattice? (iii) Is the fractional robust stable matching polytope integral? We show that all three properties hold when any number of agents on one side change preferences, as long as at most one agent on the other side does. For the case where two or more agents on both sides change preferences, we construct examples showing that both the lattice structure and polyhedral integrality fail-identifying this setting as a sharp threshold. We also present an XP-time algorithm for the general case, which implies a polynomial-time algorithm when the number of agents with changing preferences is constant. While these results establish the tractability of these regimes, closing the complexity gap in the fully general setting remains an interesting open question.

cs.DM↗

arXiv2025

Text-to-SQL Domain Adaptation via Human-LLM Collaborative Data Annotation

Text-to-SQL models, which parse natural language (NL) questions to executable SQL queries, are increasingly adopted in real-world applications. However, deploying such models in the real world often requires adapting them to the highly specialized database schemas used in specific applications. We find that existing text-to-SQL models experience significant performance drops when applied to new schemas, primarily due to the lack of domain-specific data for fine-tuning. This data scarcity also limits the ability to effectively evaluate model performance in new domains. Continuously obtaining high-quality text-to-SQL data for evolving schemas is prohibitively expensive in real-world scenarios. To bridge this gap, we propose SQLsynth, a human-in-the-loop text-to-SQL data annotation system. SQLsynth streamlines the creation of high-quality text-to-SQL datasets through human-LLM collaboration in a structured workflow. A within-subjects user study comparing SQLsynth with manual annotation and ChatGPT shows that SQLsynth significantly accelerates text-to-SQL data annotation, reduces cognitive load, and produces datasets that are more accurate, natural, and diverse. Our code is available at https://github.com/magic-YuanTian/SQLsynth.

cs.HC↗

arXiv2025

CodeLutra: Boosting LLM Code Generation via Preference-Guided Refinement

Large Language Models (LLMs) have revolutionized code generation but require significant resources and often over-generalize, limiting their task-specific efficiency. Fine-tuning smaller, open-source LLMs provides a cost-effective alternative. However, standard supervised approaches rely only on correct examples, missing valuable insights from failures. We introduce CodeLutra, a framework that leverages both correct and incorrect code attempts. Instead of using only correct solutions, CodeLutra applies iterative preference-based refinement, comparing successful and failed outputs to better approximate desired results. This approach narrows the performance gap with state-of-the-art larger models without requiring massive datasets or auxiliary models. For instance, on a challenging data science coding task, using only 500 samples improved Llama-3-8B's accuracy from 28.2% to 48.6%, approaching GPT-4's level. By learning from both successes and mistakes, CodeLutra provides a scalable and efficient path to high-quality code generation, making smaller open-source models more competitive with leading closed-source alternatives.

cs.CL↗

arXiv2025

From Documents to Dialogue: Building KG-RAG Enhanced AI Assistants

The Adobe Experience Platform AI Assistant is a conversational tool that enables organizations to interact seamlessly with proprietary enterprise data through a chatbot. However, due to access restrictions, Large Language Models (LLMs) cannot retrieve these internal documents, limiting their ability to generate accurate zero-shot responses. To overcome this limitation, we use a Retrieval-Augmented Generation (RAG) framework powered by a Knowledge Graph (KG) to retrieve relevant information from external knowledge sources, enabling LLMs to answer questions over private or previously unseen document collections. In this paper, we propose a novel approach for building a high-quality, low-noise KG. We apply several techniques, including incremental entity resolution using seed concepts, similarity-based filtering to deduplicate entries, assigning confidence scores to entity-relation pairs to filter for high-confidence pairs, and linking facts to source documents for provenance. Our KG-RAG system retrieves relevant tuples, which are added to the user prompts context before being sent to the LLM generating the response. Our evaluation demonstrates that this approach significantly enhances response relevance, reducing irrelevant answers by over 50% and increasing fully relevant answers by 88% compared to the existing production system.

cs.IR↗

arXiv2025

Detecting Ambiguities to Guide Query Rewrite for Robust Conversations in Enterprise AI Assistants

Multi-turn conversations with an Enterprise AI Assistant can be challenging due to conversational dependencies in questions, leading to ambiguities and errors. To address this, we propose an NLU-NLG framework for ambiguity detection and resolution through reformulating query automatically and introduce a new task called "Ambiguity-guided Query Rewrite." To detect ambiguities, we develop a taxonomy based on real user conversational logs and draw insights from it to design rules and extract features for a classifier which yields superior performance in detecting ambiguous queries, outperforming LLM-based baselines. Furthermore, coupling the query rewrite module with our ambiguity detecting classifier shows that this end-to-end framework can effectively mitigate ambiguities without risking unnecessary insertions of unwanted phrases for clear queries, leading to an improvement in the overall performance of the AI Assistant. Due to its significance, this has been deployed in the real world application, namely Adobe Experience Platform AI Assistant.

cs.CL↗

arXiv2024

Hallucination Diversity-Aware Active Learning for Text Summarization

Large Language Models (LLMs) have shown propensity to generate hallucinated outputs, i.e., texts that are factually incorrect or unsupported. Existing methods for alleviating hallucinations typically require costly human annotations to identify and correct hallucinations in LLM outputs. Moreover, most of these methods focus on a specific type of hallucination, e.g., entity or token errors, which limits their effectiveness in addressing various types of hallucinations exhibited in LLM outputs. To our best knowledge, in this paper we propose the first active learning framework to alleviate LLM hallucinations, reducing costly human annotations of hallucination needed. By measuring fine-grained hallucinations from errors in semantic frame, discourse and content verifiability in text summarization, we propose HAllucination Diversity-Aware Sampling (HADAS) to select diverse hallucinations for annotations in active learning for LLM finetuning. Extensive experiments on three datasets and different backbone models demonstrate advantages of our method in effectively and efficiently mitigating LLM hallucinations.

cs.CL↗

arXiv2023

Dynamic Vector Bin Packing for Online Resource Allocation in the Cloud

Several cloud-based applications, such as cloud gaming, rent servers to execute jobs which arrive in an online fashion. Each job has a resource demand and must be dispatched to a cloud server which has enough resources to execute the job, which departs after its completion. Under the `pay-as-you-go' billing model, the server rental cost is proportional to the total time that servers are actively running jobs. The problem of efficiently allocating a sequence of online jobs to servers without exceeding the resource capacity of any server while minimizing total server usage time can be modelled as a variant of the dynamic bin packing problem (DBP), called MinUsageTime DBP. In this work, we initiate the study of the problem with multi-dimensional resource demands (e.g. CPU/GPU usage, memory requirement, bandwidth usage, etc.), called MinUsageTime Dynamic Vector Bin Packing (DVBP). We study the competitive ratio (CR) of Any Fit packing algorithms for this problem. We show almost-tight bounds on the CR of three specific Any Fit packing algorithms, namely First Fit, Next Fit, and Move To Front. We prove that the CR of Move To Front is at most $(2μ+1)d +1$, where $μ$ is the ratio of the max/min item durations. For $d=1$, this significantly improves the previously known upper bound of $6μ+7$ (Kamali & Lopez-Ortiz, 2015). We then prove the CR of First Fit and Next Fit are bounded by $(μ+2)d+1$ and $2μd+1$, respectively. Next, we prove a lower bound of $(μ+1)d$ on the CR of any Any Fit packing algorithm, an improved lower bound of $2μd$ for Next Fit, and a lower bound of $2μ$ for Move To Front in the 1-D case. All our bounds improve or match the best-known bounds for the 1-D case. Finally, we experimentally study the average-case performance of these algorithms on randomly generated synthetic data, and observe that Move To Front outperforms other Any Fit packing algorithms.

cs.DS↗

arXiv2023

Optimal Sketching Bounds for Sparse Linear Regression

We study oblivious sketching for $k$-sparse linear regression under various loss functions such as an $\ell_p$ norm, or from a broad class of hinge-like loss functions, which includes the logistic and ReLU losses. We show that for sparse $\ell_2$ norm regression, there is a distribution over oblivious sketches with $Θ(k\log(d/k)/\varepsilon^2)$ rows, which is tight up to a constant factor. This extends to $\ell_p$ loss with an additional additive $O(k\log(k/\varepsilon)/\varepsilon^2)$ term in the upper bound. This establishes a surprising separation from the related sparse recovery problem, which is an important special case of sparse regression. For this problem, under the $\ell_2$ norm, we observe an upper bound of $O(k \log (d)/\varepsilon + k\log(k/\varepsilon)/\varepsilon^2)$ rows, showing that sparse recovery is strictly easier to sketch than sparse regression. For sparse regression under hinge-like loss functions including sparse logistic and sparse ReLU regression, we give the first known sketching bounds that achieve $o(d)$ rows showing that $O(μ^2 k\log(μn d/\varepsilon)/\varepsilon^2)$ rows suffice, where $μ$ is a natural complexity parameter needed to obtain relative error bounds for these loss functions. We again show that this dimension is tight, up to lower order terms and the dependence on $μ$. Finally, we show that similar sketching bounds can be achieved for LASSO regression, a popular convex relaxation of sparse regression, where one aims to minimize $\|Ax-b\|_2^2+λ\|x\|_1$ over $x\in\mathbb{R}^d$. We show that sketching dimension $O(\log(d)/(λ\varepsilon)^2)$ suffices and that the dependence on $d$ and $λ$ is tight.

cs.DS↗

arXiv2022

Sample Constrained Treatment Effect Estimation

Treatment effect estimation is a fundamental problem in causal inference. We focus on designing efficient randomized controlled trials, to accurately estimate the effect of some treatment on a population of $n$ individuals. In particular, we study sample-constrained treatment effect estimation, where we must select a subset of $s \ll n$ individuals from the population to experiment on. This subset must be further partitioned into treatment and control groups. Algorithms for partitioning the entire population into treatment and control groups, or for choosing a single representative subset, have been well-studied. The key challenge in our setting is jointly choosing a representative subset and a partition for that set. We focus on both individual and average treatment effect estimation, under a linear effects model. We give provably efficient experimental designs and corresponding estimators, by identifying connections to discrepancy minimization and leverage-score-based sampling used in randomized numerical linear algebra. Our theoretical results obtain a smooth transition to known guarantees when $s$ equals the population size. We also empirically demonstrate the performance of our algorithms.

cs.LG↗

arXiv2022

Exact Representation of Sparse Networks with Symmetric Nonnegative Embeddings

Many models for undirected graphs are based on factorizing the graph's adjacency matrix; these models find a vector representation of each node such that the predicted probability of a link between two nodes increases with the similarity (dot product) of their associated vectors. Recent work has shown that these models are unable to capture key structures in real-world graphs, particularly heterophilous structures, wherein links occur between dissimilar nodes. In contrast, a factorization with two vectors per node, based on logistic principal components analysis (LPCA), has been proven not only to represent such structures, but also to provide exact low-rank factorization of any graph with bounded max degree. However, this bound has limited applicability to real-world networks, which often have power law degree distributions with high max degree. Further, the LPCA model lacks interpretability since its asymmetric factorization does not reflect the undirectedness of the graph. We address these issues in two ways. First, we prove a new bound for the LPCA model in terms of arboricity rather than max degree; this greatly increases the bound's applicability to many sparse real-world networks. Second, we propose an alternative graph model whose factorization is symmetric and nonnegative, which allows for link predictions to be interpreted in terms of node clusters. We show that the bounds for exact representation in the LPCA model extend to our new model. On the empirical side, our model is optimized effectively on real-world graphs with gradient descent on a cross-entropy loss. We demonstrate its effectiveness on a variety of foundational tasks, such as community detection and link prediction.

cs.LG↗

arXiv2022

A Structural and Algorithmic Study of Stable Matching Lattices of "Nearby" Instances, with Applications

Recently MV18 identified and initiated work on the new problem of understanding structural relationships between the lattices of solutions of two "nearby" instances of stable matching. They also gave an application of their work to finding a robust stable matching. However, the types of changes they allowed in going from instance $A$ to $B$ were very restricted, namely any one agent executes an upward shift. In this paper, we allow any one agent to permute its preference list arbitrarily. Let $M_A$ and $M_B$ be the sets of stable matchings of the resulting pair of instances $A$ and $B$, and let $\mathcal{L}_A$ and $\mathcal{L}_B$ be the corresponding lattices of stable matchings. We prove that the matchings in $M_A \cap M_B$ form a sublattice of both $\mathcal{L}_A$ and $\mathcal{L}_B$ and those in $M_A \setminus M_B$ form a join semi-sublattice of $\mathcal{L}_A$. These properties enable us to obtain a polynomial time algorithm for not only finding a stable matching in $M_A \cap M_B$, but also for obtaining the partial order, as promised by Birkhoff's Representation Theorem, thereby enabling us to generate all matchings in this sublattice. Our algorithm also helps solve a version of the robust stable matching problem. We discuss another potential application, namely obtaining new insights into the incentive compatibility properties of the Gale-Shapley Deferred Acceptance Algorithm.

cs.DM↗

arXiv2022

Online Balanced Experimental Design

e consider the experimental design problem in an online environment, an important practical task for reducing the variance of estimates in randomized experiments which allows for greater precision, and in turn, improved decision making. In this work, we present algorithms that build on recent advances in online discrepancy minimization which accommodate both arbitrary treatment probabilities and multiple treatments. The proposed algorithms are computational efficient, minimize covariate imbalance, and include randomization which enables robustness to misspecification. We provide worst case bounds on the expected mean squared error of the causal estimate and show that the proposed estimator is no worse than an implicit ridge regression, which are within a logarithmic factor of the best known results for offline experimental design. We conclude with a detailed simulation study showing favorable results relative to complete randomization as well as to offline methods for experimental design with time complexities exceeding our algorithm.

stat.ME↗

arXiv2022

Electra: Conditional Generative Model based Predicate-Aware Query Approximation

The goal of Approximate Query Processing (AQP) is to provide very fast but "accurate enough" results for costly aggregate queries thereby improving user experience in interactive exploration of large datasets. Recently proposed Machine-Learning based AQP techniques can provide very low latency as query execution only involves model inference as compared to traditional query processing on database clusters. However, with increase in the number of filtering predicates(WHERE clauses), the approximation error significantly increases for these methods. Analysts often use queries with a large number of predicates for insights discovery. Thus, maintaining low approximation error is important to prevent analysts from drawing misleading conclusions. In this paper, we propose ELECTRA, a predicate-aware AQP system that can answer analytics-style queries with a large number of predicates with much smaller approximation errors. ELECTRA uses a conditional generative model that learns the conditional distribution of the data and at runtime generates a small (~1000 rows) but representative sample, on which the query is executed to compute the approximate result. Our evaluations with four different baselines on three real-world datasets show that ELECTRA provides lower AQP error for large number of predicates compared to baselines.

cs.DB↗

arXiv2021

Online MAP Inference and Learning for Nonsymmetric Determinantal Point Processes

In this paper, we introduce the online and streaming MAP inference and learning problems for Non-symmetric Determinantal Point Processes (NDPPs) where data points arrive in an arbitrary order and the algorithms are constrained to use a single-pass over the data as well as sub-linear memory. The online setting has an additional requirement of maintaining a valid solution at any point in time. For solving these new problems, we propose algorithms with theoretical guarantees, evaluate them on several real-world datasets, and show that they give comparable performance to state-of-the-art offline algorithms that store the entire data in memory and take multiple passes over it.

cs.LG↗

arXiv2021

Optimal Space and Time for Streaming Pattern Matching

In this work, we study longest common substring, pattern matching, and wildcard pattern matching in the asymmetric streaming model. In this streaming model, we have random access to one string and streaming access to the other one. We present streaming algorithms with provable guarantees for these three fundamental problems. In particular, our algorithms for pattern matching improve the upper bound and beat the unconditional lower bounds on the memory of randomized and deterministic streaming algorithms. In addition to this, we present algorithms for wildcard pattern matching in the asymmetric streaming model that have optimal space and time.

cs.DS↗