--- title: Trading, Microstructure & Execution url: https://www.ml-quant.com/topics/trading-microstructure-execution/ site: ML-Quant (https://www.ml-quant.com) updated: 2026-10-09 license: Summaries CC BY 4.0; links go to the original sources index: https://www.ml-quant.com/llms.txt --- # Trading, Microstructure & Execution Order books, market making, execution, high-frequency data and trading signals. 559 papers featured; 45 in the last 12 months. Papers featured per quarter: 2023 Q2 27, 2023 Q3 81, 2023 Q4 62, 2024 Q1 71, 2024 Q2 77, 2024 Q3 51, 2024 Q4 31, 2025 Q1 52, 2025 Q2 46, 2025 Q3 15, 2025 Q4 14, 2026 Q1 1, 2026 Q2 0, 2026 Q3 10, 2026 Q4 21 ## Most cited - [FAST: Efficient Action Tokenization for Vision-Language-Action Models](https://www.ml-quant.com/papers/arxiv/2501.09747/): 741 citations. A new tokenization scheme, Frequency-space Action Sequence Tokenization (FAST), has been proposed for robot actions, facilitating the training of vision-language action policies for complex and high-frequency tasks. - [Deep Reinforcement Learning for Active High Frequency Trading](https://www.ml-quant.com/papers/arxiv/2101.07107/): 52 citations. A new Deep Reinforcement Learning framework has been developed for high frequency stock trading, showing potential for profitable long-term strategies. - [Transformers versus LSTMs for electronic trading](https://www.ml-quant.com/papers/arxiv/2309.11400/): 33 citations. A comparison study of LSTM-based and Transformer-based models for financial time series prediction found that LSTM-based models perform better in difference sequence prediction, despite Transformer-based models having limited advantages in absolute price sequence prediction. - [JAX-LOB: A GPU-Accelerated limit order book simulator to unlock large scale reinforcement learning for trading](https://www.ml-quant.com/papers/arxiv/2308.13289/): 32 citations. JAX-LOB: The paper introduces JAX-LOB, the first GPU-powered limit order book simulator capable of processing multiple books simultaneously, designed for efficient large-scale simulations of LOB dynamics for research, calibration, and reinforcement learning training. - [Deep attentive survival analysis in limit order books: estimating fill probabilities with convolutional-transformers](https://www.ml-quant.com/papers/arxiv/2306.05479/): 28 citations. A deep learning method outperforms other approaches in estimating filltimes of limit orders. - [StockGPT: A GenAI Model for Stock Prediction and Trading](https://www.ml-quant.com/papers/arxiv/2404.05101/): 27 citations. The study introduces StockGPT, a model that predicts stock return dynamics using AI, showcasing the potential of AI in complex financial investment decisions. - [Deep Reinforcement Learning for Quantitative Trading](https://www.ml-quant.com/papers/arxiv/2312.15730/): 27 citations. AI and machine learning are revolutionizing quantitative trading with advanced algorithms, including a new model, QTNet, that uses deep reinforcement learning to manage volatile financial data. - [Conditional Generators for Limit Order Book Environments: Explainability, Challenges, and Robustness](https://www.ml-quant.com/papers/arxiv/2306.12806/): 27 citations. Conditional generative models used for order book simulation, enhanced with adversarial attacks. - [E-backtesting](https://www.ml-quant.com/papers/arxiv/2209.00991/): 24 citations. A new backtesting procedure for Expected Shortfall forecasts is proposed using e-values and e-processes. - [HLOB - Information Persistence and Structure in Limit Order Books](https://www.ml-quant.com/papers/arxiv/2405.18938/): 22 citations. A new deep learning model, HLOB, has been developed for predicting Limit Order Book mid-price changes, outperforming nine other models and offering new insights into information distribution in Limit Order Books. - [Don't Let MEV Slip: The Costs of Swapping on the Uniswap Protocol](https://www.ml-quant.com/papers/arxiv/2309.13648/): 21 citations. Efficiency and Slippage Analysis: The research analyzes the costs of trading on a decentralized exchange, showing that costs vary based on trade characteristics, and proposes that DEXs could be a trustworthy alternative to centralized exchanges for trading digital assets. - [Strict Universality of the Square-Root Law in Price Impact across Stocks: A Complete Survey of the Tokyo Stock Exchange.](https://www.ml-quant.com/papers/arxiv/2411.13965/): 20 citations. Research using Tokyo Stock Exchange data supports the econophysics theory that average price impact follows a power law in relation to transaction volume. ## Latest - [Event History Over Scale: Compact Transformers for Low-Latency Limit Order Book Forecasting](https://www.ml-quant.com/papers/arxiv/2610.02917/) (2026-10-09): MBOFormer and MBOFusion, two causal transformers with under 70 kilobyte serialized state, predict limit order book price trends with sub-millisecond latency across three markets and four prediction horizons. - [Latent Continuum of Regimes in Limit Order Book Dynamics](https://www.ml-quant.com/papers/arxiv/2610.05740/) (2026-10-09): Analysis of high-frequency limit-order-book geometry finds market states form a continuous structure with a dominant latent coordinate capturing between 83.66% and 85.27% of covariance variation. - [How Execution Assumptions Change Short-Horizon Sharpe Rankings: Evidence from a Synthetic Trading Benchmark](https://www.ml-quant.com/papers/arxiv/2610.05077/) (2026-10-09): Varying execution realism from ideal fills to latency and impact reshuffles rankings of LLM and classical trading policies, showing how backtest conventions affect headline results. - [Modelling Regime Shifts in Continuous Intraday Electricity Markets with State-dependent Hawkes Processes](https://www.ml-quant.com/papers/arxiv/2610.08169/) (2026-10-09): A regime-switching Hawkes process models intraday electricity order flow across liquidity conditions, showing that stress mainly strengthens self-excitation rather than reshaping cross-side dynamics. - [What Survives Honest Tes1ng? A Pre-registered Test of Eleven Trading Signals](https://www.ml-quant.com/papers/ssrn/7553278/) (2026-10-09): Applies a strict pre-registered protocol to eleven trading signals across equity and futures markets, finding only trend survives development and hold-out periods with borderline significance after multiple-testing adjustment. - [The Toll-Fenced Market: Documented Anomalies in Indian Equities Replicate but do Not Survive Retail Implementation Costs](https://www.ml-quant.com/papers/ssrn/7551520/) (2026-10-09): Well-documented Indian equity anomalies replicate strongly in-sample but all fail to survive 12 to 32 basis point retail transaction costs; required directional accuracy reaches 72% at hourly horizons. - [Certified Alpha Capacity: Statistical Evidence, Economic Lifetime, and Arbitrage under Decay](https://www.ml-quant.com/papers/arxiv/2610.01115/) (2026-10-02): The research measures when a trading signal accumulates enough statistical evidence for deployment before its economic value decays, deriving exact feasibility thresholds. - [Evolution of Market Microstructure in the Age of AI](https://www.ml-quant.com/papers/arxiv/2609.33058/) (2026-10-02): A survey traces how market design has evolved from floor trading through electronic exchanges to AI agents, examining whether markets can allocate goods efficiently without full revelation. - [Say, Echo, Do: Strategic Narratives and Revealed Positioning in Financial Markets](https://www.ml-quant.com/papers/arxiv/2609.38545/) (2026-10-02): The research shows when to follow or fade financial narratives by studying the covariance between institutional statements and revealed trading positions using machine learning. - [UQ-LOB: Uncertainty-Aware Limit Order Book Mid-Price Forecasting](https://www.ml-quant.com/papers/arxiv/2609.31491/) (2026-10-02): A lightweight module adds confidence estimates to limit-order-book forecasters, raising directional F1 by 0.11-0.15 when predicting the most confident 10% of trades. - [ReLOBGen: Replayable Limit Order Book Message Generation](https://www.ml-quant.com/papers/arxiv/2609.35867/) (2026-10-02): A method generates realistic limit order book messages guaranteed to be consistent with current market state, achieving 100% replayability and 2.7–3.6× speedup over existing approaches. - [Packets, Transactions and Queues: Design Principles for HFT Systems from a Measurement Study of CME Market Data](https://www.ml-quant.com/papers/arxiv/2609.32848/) (2026-10-02): Measurement of over a year of CME market data reveals transactions cluster within microseconds, yielding design principles: single-thread receivers never queue, two-thread splits reduce tail latency. - [PPO-HRAP: Proximal Policy Optimization with a Hybrid Regime-Aware Policy for Risk-Controlled Trading](https://www.ml-quant.com/papers/arxiv/2610.01325/) (2026-10-02): Reinforcement learning agent combining policy optimization with regime priors controls drawdown while blending learned and rule-based portfolio exposure on held-out equity data. - [Agentic Limit Order Books: Phase Transitions and Market Impact](https://www.ml-quant.com/papers/arxiv/2609.31260/) (2026-10-02): The study shows that reinforcement-learning traders in limit order books create phase transitions between orderly trading and volatile cascades, with non-linear market impact dynamics. - [Order Flow Imbalance and Price Discovery on Prediction Markets: Evidence from Kalshi](https://www.ml-quant.com/papers/ssrn/7520438/) (2026-10-02): Order flow imbalance predicts contemporaneous mid-price changes in Kalshi binary event contracts, with explanatory power varying from 0.29 for sports to 0.02 for macroeconomic events. - [Spot Bitcoin ETF Flows and Bitcoin Returns: Timestamp Alignment and Asymmetric Feedback](https://www.ml-quant.com/papers/ssrn/7545019/) (2026-10-02): Correcting for timestamp mismatch between ETF and Bitcoin markets, the study finds that same-day flows predict next-day returns at 1.67 percentage points per billion of inflow, while prior Bitcoin returns predict flows. - [Does Crypto Trading Activity Follow New York Time? Cross-Venue and New York-London Evidence from Clock Changes](https://www.ml-quant.com/papers/ssrn/7526338/) (2026-10-02): Bitcoin and Ether trading activity on major venues increases during the UTC window aligned with New York market open, with the effect appearing on both assets. - [Evolution of Market Microstructure in the Age of AI From Double-Sided Continuous Auctions to Agentic Markets](https://www.ml-quant.com/papers/ssrn/7530158/) (2026-10-02): A survey traces how market design rules evolved from floor trading to limit order books to automated market makers to agent-based trading systems. - [Market Attention and Information Quality in On-Chain Prediction Markets: Evidence from Polymarket](https://www.ml-quant.com/papers/ssrn/7540819/) (2026-10-02): On-chain prediction market analysis finds 68.2 percent of volume sits in well-calibrated markets with average seven-day pricing error of 0.03. - [Order Flow Imbalance and Price Impact in Indian Equity Derivatives: Cross-Sectional Evidence from NSE F&O Markets](https://www.ml-quant.com/papers/ssrn/7542098/) (2026-10-02): Multilevel order flow imbalance in Indian equity derivatives predicts mean-reverting price dynamics with information coefficients growing from -0.004 at ten seconds to -0.032 at sixty seconds. - [One dollar, many prices: dealer-specific pricing of synthetic dollar funding](https://www.ml-quant.com/papers/repec/boe-boeewp-023631/) (2026-10-02): Comparing FX forwards in identical currency pairs and maturities, the study finds large pricing variation across dealers reflecting clientele and pricing power rather than funding costs. - [Robust Market Making with Hawkes Order Flow and Price Impact via Adversarial Reinforcement Learning](https://www.ml-quant.com/papers/arxiv/2609.22785/) (2026-09-25): The research extends adversarial reinforcement learning for market making to handle self-exciting order arrivals and price impact, using an LSTM module to improve robustness in complex microstructure environments. - [Liquidity Provision and Rebate Design in Option Markets](https://www.ml-quant.com/papers/arxiv/2609.26606/) (2026-09-25): Develops a nested optimization model for market making and rebate design in option markets, showing how exchanges can set fees to incentivize liquidity provision and improve market depth. - [Feasible Multi-Asset Optimal Execution under Cash Constraints](https://www.ml-quant.com/papers/arxiv/2609.27786/) (2026-09-25): Extends the Almgren-Chriss optimal execution framework to enforce intertemporal cash constraints, reducing peak cash drawdown while maintaining implementation shortfall in multi-asset rebalancing. - [Rule-Based Pricing Algorithms and Market Outcomes: An Experimental Study](https://www.ml-quant.com/papers/arxiv/2609.26861/) (2026-09-25): Experiments show that algorithm design features like warnings, pre-configured strategies, and LLM advice raise market prices by increasing starting prices and fostering cooperative algorithm designs. - [Algorithmic Collusion by Reinforcement-Learning Pricing Agents: Simulation Evidence and Implications for Financial Markets and Competition Law](https://www.ml-quant.com/papers/ssrn/7500483/) (2026-09-25): Q-learning pricing agents in simulated duopolies reach supracompetitive outcomes with no communication, achieving collusion indices of 0.778 and 40% profit gains over competitive benchmarks. - [Incentives at Play: Fee-Induced Volume on a Regulated Perpetual Futures Venue](https://www.ml-quant.com/papers/ssrn/7512338/) (2026-09-25): Analysis of Kalshi's regulated Bitcoin and Ethereum futures reveals that 39-48% of notional trades are mechanical fixed-size orders that vanish when fees are charged, indicating costless artificial volume rather than legitimate trading. - [Settlement Risk and Currency Markets](https://www.ml-quant.com/papers/ssrn/7201787/) (2026-09-25): Hungary's 2015 adoption of payment-versus-payment settlement reduced currency excess returns by ten basis points, demonstrating settlement risk is a priced friction limiting arbitrage. - [Fedspeak, LLM-Derived Signals, and High-Frequency Trading](https://www.ml-quant.com/papers/ssrn/7479619/) (2026-09-25): Semantic and tonal shifts across sequential Federal Reserve communications generate significant intraday price movements and abnormal volume, revealing incomplete information absorption at initial announcement. - [Elastic in cash, inelastic in repo: Hedge funds in the treasury and repo markets](https://www.ml-quant.com/papers/repec/zbw-safewp-343098/) (2026-09-25): Using German sovereign bond repo data, the research shows hedge funds are price-elastic in cash markets but highly inelastic in repo, inheriting elasticity from their cash-market counterparties.