Publications
Filter by area, type, author, year, or venue — or just type a few words.
Search publications
Uncertainty Quantification of Click and Conversion Estimates for the Autobidding
We propose the DenoiseBid method, which corrects the generated CTRs and CVRs to make the resulting bids more efficient in auctions.
UCB-Based Feature Engineering for Cold-Start in Recommenders
We introduce UCB-FE, a model-agnostic technique that transforms behavioral features with position-aware upper confidence bounds, boosting cold items in CTR ranking without any weight or architecture changes.
Fast and Feasible: Permutation-based Constrained Reranking for Revenue Maximization
We state the reranking problem as an integer linear program (ILP) that maximizes revenue subject to per-query constraints on other metrics, e.g., relevance.
RRE-GTC: A Geo-Temporal Cluster Dataset for Real-Estate PriceEstimation and Market-Aware Search
We introduce RRE-GTC (Ru-Real-Estate Geo-Temporal Clusters), an open resource designed to benchmark ranking, pricing, and recommendation algorithms in dynamic spatial contexts.
Practical Training-Free MCTS Query Optimization
This paper presents MCTS-Extreme, a training-free Monte Carlo Tree Search join optimizer that reuses PostgreSQL's own cost model and runs as a native planner extension, beating GEQO and DPSize on high-arity queries. A reproduction study finds the learned MCTS systems we could run did not beat the default optimizer.
Training-Free Query Optimization via LLM-Based Plan Similarity
We introduce LLM-based Plan Mapping, a framework that embeds the default execution plan of a query, finds its k nearest neighbors among previously executed plans, and recommends database hintsets based on neighborhood voting.
Functional multi-armed bandit and the best function identification problems
We propose two new classes of problems: the functional multi-armed bandit problem (FMAB) and the best function identification problem.
Robust autobidding for noisy conversion prediction models
We propose RobustBid, an efficient method for robust autobidding taking into account uncertainty in CTR and CVR predictions.
A duration-dependent flood depth-damage function calibrated using FEMA NFIP claims from three U.S. hurricanes
We calibrate a duration-dependent flood depth-damage function on 317,943 FEMA NFIP claims from three U.S. hurricanes, showing prolonged flooding raises damage up to 2.6× at the same depth and enabling integration into banks' credit risk models.
VK-LSVD: A Large-Scale Industrial Dataset for Short-Video Recommendation
We introduce the VK Large Short-Video Dataset (VK-LSVD), the largest publicly available industrial dataset of its kind.
Survey of Modern Smooth Optimization Algorithms with Comparison Oracle
This review provides an overview of contemporary algorithms for smooth, multivariate optimization that utilize only information about the order of the function values, rather than their numerical magnitudes.
Zeroth-order methods for non-smooth stochastic problems under heavy-tailed noise
We propose gradient-free algorithms with zeroth-order oracle under adversarial noise with unbounded variance, for non-smooth convex and convex-concave optimization problems.
Identification of the Braess paradox in a stable dynamic model in network with one source and multiple sinks
We study the problem of identifying edges in a transportation graph where the introduction of an additional toll would enhance the efficiency of network usage within the Nesterov–de Palma equilibrium model.
Hint Based Query Optimization with LLM Agent and Plan Similarity
A two-path architecture for query-optimizer hint selection that combines fast nearest-neighbor transfer in an LLM-derived plan-embedding space with a budgeted LLM agent that searches the hint space under DBMS feedback.
UCB-type Algorithm for Budget-Constrained Expert Learning
We introduce M-LCB, a computationally efficient UCB-style meta-algorithm that provides anytime regret guarantees.
Autobidding Arena: unified evaluation of the classical and RL-based autobidding algorithms
We consider the most efficient autobidding algorithms from different classes, e.g., ones based on the controllers, RL, optimal formulas, etc., and benchmark them in the bidding environment.
RARe: Raising Ad Revenue Framework with Context-Aware Reranking
We propose and compare two different click models that take into account the context of items in a search result.
Fast UCB-type algorithms for stochastic bandits with heavy and super heavy symmetric noise
We propose a new method for constructing UCB-type algorithms for stochastic multi-armed bandits based on general convex optimization methods with an inexact oracle.
Bat: Benchmark for auto-bidding task
An open benchmark and dataset for real-time ad-auction bidding, with baselines for budget pacing and cost-per-click constraints, so autobidding algorithms can be developed and compared on common ground.
Optimizing Online Advertising with Multi-Armed Bandits: Mitigating the Cold Start Problem under Auction Dynamics
We developed a UCB-like algorithm under multi-armed bandit (MAB) setting for positional-based model (PBM), specifically tailored to auction pay-per-click systems.
Optimal Traffic Allocation for Multi-Slot Sponsored Search: Balance of Efficiency and Fairness
We propose a novel ad allocation model that departs from traditional auction mechanics.
Power of Generalized Smoothness in Stochastic Convex Optimization: First- and Zero-Order Algorithms
This paper is devoted to the study of stochastic optimization problems under the generalized smoothness assumption.
γ-Competitiveness: An Approach to Multi-Objective Optimization with High Computation Costs in Lipschitz Functions
We introduce an extension of the concept of competitive solutions and propose the Scalarization With Competitiveness Method (SWCM) for multi-criteria problems.
On quasi-convex smooth optimization problems by a comparison oracle
This paper is devoted to an approach to minimizing quasi-convex functions using a recently proposed comparison oracle only.
Learning-Augmented Online Caching: New Upper Bounds
We address the problem of learning-augmented online caching for DBMS in the scenario when each request is accompanied by a prediction of the next occurrence of the requested page.
EEvA: Fast Expert-Based Algorithms for Buffer Page Replacement
In this paper, we propose a new family of page replacement algorithms for DB buffer manager which demonstrate a superior performance wrt competitors on custom data access patterns and imply a low computational overhead on TPC-C.
Implicitly normalized forecaster with clipping for linear and non-linear heavy-tailed multi-armed bandits
In this paper, we propose a new version of INF called the Implicitly Normalized Forecaster with clipping (INF-clip) for MAB problems with heavy-tailed reward distributions.