arXiv subjects

Aleksei Smirnov

Publications and source records attributed to Aleksei Smirnov.

6 recordsLinked to original sources

arXiv2026

The Bidding Games: Reinforcement Learning for MEV Extraction on Polygon Blockchain

In blockchain networks, the strategic ordering of transactions within blocks has emerged as a significant source of profit extraction, known as Maximal Extractable Value (MEV). The transition from spam-based Priority Gas Auctions (PGA) to structured auction mechanisms like Polygon Atlas has transformed MEV extraction from public bidding wars into sealed-bid competitions under extreme time constraints. While this shift reduces network congestion, it introduces complex strategic challenges where searchers must make optimal bidding decisions within a sub-second window without knowledge of competitor behavior or presence. Traditional equilibrium-based game-theoretic models struggle in this high-frequency, partially observable environment. While auction theory provides equilibrium solutions for sealed-bid formats under incomplete information, these models typically assume known bidder value distributions and stationary competition--assumptions that are difficult to satisfy in dynamic, sub-second auctions where competitor presence and strategies evolve rapidly. We present a reinforcement learning framework for MEV extraction on Polygon Atlas and make three contributions: (1) A novel simulation environment that accurately models the stochastic arrival of arbitrage opportunities and probabilistic competition in Atlas auctions; (2) A PPO-based bidding agent optimized for real-time constraints, capable of adaptive strategy formulation in continuous action spaces while maintaining production-ready inference speeds; (3) Empirical validation demonstrating our history-conditioned agent achieves 49\% Maximum-Profit Capture when deployed alongside existing searchers and a 43\% relative profit improvement over the historical market leader in counterfactual replacement, significantly outperforming static bidding strategies.

cs.GT↗

arXiv2026

The Origins of MEV: Systematic Attribution of Arbitrage Opportunity Creation at Scale

Maximal Extractable Value (MEV) represents billions of dollars in extracted value that fundamentally shapes blockchain network dynamics and participant incentives. While research has focused on MEV extraction and mitigation, we lack systematic methods to attribute MEV opportunities to their on-chain origins. This paper formalizes the MEV opportunity attribution problem and introduces a systems framework for identifying which transactions create arbitrage opportunities and quantifying their contributions. We design and evaluate four attribution methods for atomic arbitrage on EVM-compatible networks: bot-data-driven, simulation-based, coefficient-based, and Shapley-based approaches. Through large-scale retrospective analysis spanning over one million blocks on Polygon, we demonstrate that the majority of atomic arbitrage opportunities can be traced to single source transactions, validating our central hypothesis about competitive MEV markets. We quantify a highly concentrated distribution of MEV creation, where a small subset of protocols generates most opportunities, and provide comparative analysis of method trade-offs in accuracy, cost, and scalability. Our findings offer insights for protocol designers reducing MEV leakage, validators optimizing transaction ordering, and analysts measuring ecosystem health through opportunity creation.

cs.DC↗

arXiv2026

AI-based approach to burnout identification from textual data

This study introduces an AI-based methodology that utilizes natural language processing (NLP) to detect burnout from textual data. The approach relies on a RuBERT model originally trained for sentiment analysis and subsequently fine-tuned for burnout detection using two data sources: synthetic sentences generated with ChatGPT and user comments collected from Russian YouTube videos about burnout. The resulting model assigns a burnout probability to input texts and can be applied to process large volumes of written communication for monitoring burnout-related language signals in high-stress work environments.

cs.CL↗

arXiv2025

Crystal Eye: all sky MeV monitor with high precision real-time localization

Crystal Eye is a space-based all-sky monitor optimized for the autonomous detection and localization of transients in the 10 keV to 30 MeV energy range, a region where extensive observations and monitoring of various astrophysical phenomena are required. By focusing on the operating environment and its impact on the observation process, we optimized the detector design and assessed its scientific potential. We explored the use of novel techniques to achieve the science goals of the experiment. We assumed the orbit of a potential future mission at approximately 550 km altitude near the equatorial region with a 20° inclination. In such an orbit, the main background contributions for this kind of detector are from different particles and radiation of cosmic origin and secondaries produced by their interaction in the Earth's atmospheric and geomagnetic environment. We studied the response of the detector in this background environment, using the Geant4 Monte Carlo simulation toolkit. We also calculated other detector performance parameters to estimate its scientific capabilities. The detector effective area and efficiency are calculated for low energy gamma-ray sources and used to estimate its sensitivity to short-duration transient sources. The calculation shows a better effective area and sensitivity by several factors compared to existing instruments of similar type. A method is also developed and discussed to estimate the online transient-localization performance of the detector, suggesting a better localization precision by about an order of magnitude than those typically reported by existing gamma-ray monitors. We present here the simulation study and results of an innovative detector design concept that can make a significant contribution in the multi-messenger era. Moreover, this study can be useful as a technical reference for similar future experiments.

astro-ph.IM↗

arXiv2025

Designing Social Learning

This paper studies strategic communication in the context of social learning. Product reviews are used by consumers to learn product quality, but in order to write a review, a consumer must be convinced to purchase the item first. When reviewers care about welfare of future consumers, this leads to a conflict: a reviewer today wants the future consumers to purchase the item even when this comes at a loss to them, so that more information is revealed for the consumers that come after. We show that due to this conflict, communication via reviews is inevitably noisy, regardless of whether reviewers can commit to a communication strategy or have to resort to cheap talk. The optimal communication mechanism involves truthful communication of extreme experiences and pools the moderate experiences together.

econ.TH↗

arXiv2024

Comparative Analysis of Encoder-Based NER and Large Language Models for Skill Extraction from Russian Job Vacancies

The labor market is undergoing rapid changes, with increasing demands on job seekers and a surge in job openings. Identifying essential skills and competencies from job descriptions is challenging due to varying employer requirements and the omission of key skills. This study addresses these challenges by comparing traditional Named Entity Recognition (NER) methods based on encoders with Large Language Models (LLMs) for extracting skills from Russian job vacancies. Using a labeled dataset of 4,000 job vacancies for training and 1,472 for testing, the performance of both approaches is evaluated. Results indicate that traditional NER models, especially DeepPavlov RuBERT NER tuned, outperform LLMs across various metrics including accuracy, precision, recall, and inference time. The findings suggest that traditional NER models provide more effective and efficient solutions for skill extraction, enhancing job requirement clarity and aiding job seekers in aligning their qualifications with employer expectations. This research contributes to the field of natural language processing (NLP) and its application in the labor market, particularly in non-English contexts.

cs.CL↗