--- title: ML & AI Methods url: https://www.ml-quant.com/topics/ml-ai-methods/ site: ML-Quant (https://www.ml-quant.com) updated: 2026-09-26 license: Summaries CC BY 4.0; links go to the original sources index: https://www.ml-quant.com/llms.txt --- # ML & AI Methods Machine-learning methods applied to finance: deep learning, boosting, RL and new architectures. 1107 papers featured; 41 in the last 12 months. Papers featured per quarter: 2023 Q2 27, 2023 Q3 55, 2023 Q4 88, 2024 Q1 79, 2024 Q2 154, 2024 Q3 202, 2024 Q4 168, 2025 Q1 119, 2025 Q2 146, 2025 Q3 28, 2025 Q4 27, 2026 Q1 2, 2026 Q2 1, 2026 Q3 11 ## Most cited - [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://www.ml-quant.com/papers/arxiv/2312.00752/): 9205 citations. Sequence Modeling: Mamba, a neural network architecture that doesn't use attention or MLP blocks, provides faster inference and better performance in language, audio, and genomics than Transformers. - [Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality](https://www.ml-quant.com/papers/arxiv/2405.21060/): 1953 citations. The research identifies a link between state-space models and Transformers in deep learning, leading to the creation of a faster language modeling architecture, Mamba-2. - [Octo: An Open-Source Generalist Robot Policy](https://www.ml-quant.com/papers/arxiv/2405.12213/): 1880 citations. Octo is a large transformer-based policy for robotic manipulation, trained on a vast dataset, that can be instructed via language or images and adapted to new domains. - [Mastering Diverse Domains through World Models](https://www.ml-quant.com/papers/arxiv/2301.04104/): 1418 citations. Algorithm Mastery: DreamerV3, a universal algorithm, excels in over 150 varied tasks, including diamond collection in Minecraft without human input, expanding the scope of reinforcement learning. - [SimPO: Simple Preference Optimization with a Reference-Free Reward](https://www.ml-quant.com/papers/arxiv/2405.14734/): 1173 citations. Simple Preference Optimization: SimPO improves reinforcement learning from human feedback by using the average log probability of a sequence as the implicit reward, enhancing training stability and computational efficiency. - [Visual Autoregressive Modeling: Scalable Image Generation via Next-Scale Prediction](https://www.ml-quant.com/papers/arxiv/2404.02905/): 1169 citations. The article discusses Visual AutoRegressive modeling (VAR), a new image learning method that outperforms diffusion transformers in terms of speed, image quality, and scalability. - [PixArt-α: Fast Training of Diffusion Transformer for Photorealistic Text-to-Image Synthesis](https://www.ml-quant.com/papers/arxiv/2310.00426/): 967 citations. PIXART-$\alpha$, a Transformer-based text-to-image model, generates high-quality images at a low cost, reducing CO2 emissions and offering a cost-effective solution for the AIGC community. - [Fourier Neural Operator with Learned Deformations for PDEs on General Geometries](https://www.ml-quant.com/papers/arxiv/2207.05209/): 727 citations. The study introduces geo-FNO, a new framework for solving partial differential equations on any geometry, proving to be faster and more accurate than standard and machine learning-based solvers. - [LightGaussian: Unbounded 3D Gaussian Compression with 15x Reduction and 200+ FPS](https://www.ml-quant.com/papers/arxiv/2311.17245/): 660 citations. LightGaussian is a new method that converts 3D Gaussians into a more compact format, enhancing efficiency in real-time neural rendering and reducing storage needs. - [Gated Linear Attention Transformers with Hardware-Efficient Training](https://www.ml-quant.com/papers/arxiv/2312.06635/): 539 citations. Efficient Training of Gated Linear Attention Transformers: The research introduces a more hardware-efficient version of gated linear attention Transformers that performs well against other models, especially in training on longer sequences. - [CAT3D: Create Anything in 3D with Multi-View Diffusion Models](https://www.ml-quant.com/papers/arxiv/2405.10314/): 482 citations. Multi-View Diffusion Models: CAT3D is a novel technique for generating 3D scenes from any number of images, surpassing existing methods in speed and efficiency. - [Capabilities of Gemini Models in Medicine](https://www.ml-quant.com/papers/arxiv/2404.18416/): 446 citations. Med-Gemini, an AI model for medical applications, outperforms previous models and human experts in medical benchmarks, indicating potential for real-world medical use. ## Latest - [Propose, Don't Judge: An Anytime-Valid Referee for LLM Agents That Mine Investment Factors](https://www.ml-quant.com/papers/arxiv/2609.27051/) (2026-09-25): Proposes a statistical referee that judges investment factors proposed by language-model agents using out-of-sample market outcomes, ensuring false-discovery control at any stopping time. - [AlphaDiverse: Post-Training Local Quantitative Research Agents for Diverse Exploration in Alpha Factor Mining](https://www.ml-quant.com/papers/arxiv/2609.29014/) (2026-09-25): Proposes a multi-agent system with post-training that automates alpha factor mining locally, using diverse research paths and joint optimization to broaden exploration while maintaining prediction quality. - [Forecast-Dojo: Replayable Environments for Benchmarking and Training LLM Forecasting Agents](https://www.ml-quant.com/papers/arxiv/2609.28876/) (2026-09-25): Introduces a replayable environment combining 1,568 resolved prediction-market questions with 18.8M dated news articles to benchmark and train language-model forecasting agents on historical data. - [Artificial intelligence and financial markets](https://www.ml-quant.com/papers/ssrn/7515878/) (2026-09-25): A survey examines how AI transforms information production, intermediation, and market structure, with implications for efficiency, competition and financial stability. - [Label alchemy: Target engineering for improved stock selection](https://www.ml-quant.com/papers/ssrn/7494298/) (2026-09-25): Reshaping the prediction target through location, scale and shape transformations raises long-short Sharpe from 0.68 to 1.69, with label choice mattering more than model choice. - [Memorisation or Alpha? Detecting Look-Ahead Contamination in Cross-Sectional Equity Signals](https://www.ml-quant.com/papers/ssrn/7490302/) (2026-09-25): Testing whether a large language model ranks stocks by forecasting or memory, the study finds a significant information-coefficient gap of 0.185 inside versus outside its training window, suggesting substantial look-ahead contamination. - [Beta Recall, Alpha Recall, and a Contamination Detector that Needs No Labels * Measuring Training-data Leakage in LLM Equity Signals](https://www.ml-quant.com/papers/ssrn/7485818/) (2026-09-25): The study measures recall versus forecasting in an LLM's stock rankings by comparing cross-sectional information coefficients inside and outside the training window. - [Execution-Aware Alpha Mining: Teaching LLM Factor Agents to Account for Trading Costs](https://www.ml-quant.com/papers/ssrn/7498983/) (2026-09-25): The paper builds a closed-loop system where an LLM proposes equity factors penalized for execution costs and shows that accounting for trading costs dramatically improves net performance. - [Assessing the Benefits of Optimized Agentic AI Systems for Asset Pricing](https://www.ml-quant.com/papers/repec/nbr-nberwo-35431/) (2026-09-25): Optimized AI systems analyzing earnings call transcripts double explained variation in stock returns versus standard benchmarks while improving interpretability through human-readable decision rules. - [Predicting Financial Market Stress with Machine Learning](https://www.ml-quant.com/papers/repec/cpr-ceprdp-20439/) (2026-09-25): Tree-based machine learning models predict the full distribution of financial market stress 27% better than traditional time-series methods, with macro uncertainty and monetary policy expectations as key drivers. - [Ex Machina: Financial Stability in the Age of Artificial Intelligence](https://www.ml-quant.com/papers/repec/cpr-ceprdp-20681/) (2026-09-25): Q-learning and large language model investors generate systematically different behaviors in fund redemption settings, with Q-learning showing excessive coordination and amplified fragility under default risk. - [Economics of Human and AI Collaboration: When is Partial Automation More Attractive than Full Automation?](https://www.ml-quant.com/papers/arxiv/2603.29121/) (2026-04-03): The paper discusses a model that suggests combining human effort with partial automation is usually cheaper and more effective than fully automating complex tasks. - [The Inference Bottleneck: Antitrust and Neutrality Duties in the Age of Cognitive Infrastructure](https://www.ml-quant.com/papers/arxiv/2602.22750/) (2026-03-04): The article explains that the rise of generative AI is changing competition by focusing on ongoing decision-making, which can lead to unfair practices in the market. It suggests a solution called Neutral Inference to promote fairness and transparency. - [Bounded Normative Equivalence in Human-AI Cooperation: Group Behaviour, Not Partner Labels, Predicts Cooperation under Anonymous Aggregate Feedback](https://www.ml-quant.com/papers/arxiv/2601.20487/) (2026-02-02): AI agents in group settings can influence cooperation just like humans do, showing that social norms can adjust for both AI and human participants. - [Gingado: A Machine Learning Library Focused on Economics and Finance](https://www.ml-quant.com/papers/ssrn/4482553/) (2025-12-28): ML for Economics: Gingado is a developing Python library that helps incorporate machine learning into economic research by enhancing datasets and evaluating models. - [Decoding the Unique Price Behavior in the Japanese Stock Market with Convolutional Neural Networks](https://www.ml-quant.com/papers/ssrn/4478013/) (2025-12-28): Analyzing Japanese stock charts with CNN reveals predictive patterns for returns, independent of common momentum trends. - [The Banker in Your Social Network](https://www.ml-quant.com/papers/ssrn/4466139/) (2025-12-28): The research indicates that social financial advice significantly boosts stock market participation, especially through close social ties. - [SigMA: Path Signatures and Multi-head Attention for Learning Parameters in fBm-driven SDEs](https://www.ml-quant.com/papers/arxiv/2512.15088/) (2025-12-19): The SigMA neural architecture improves parameter estimation in stochastic differential equations using path signatures and self-attention, outperforming traditional methods across multiple datasets. - [Deep Learning and Elicitability for McKean-Vlasov FBSDEs With Common Noise](https://www.ml-quant.com/papers/arxiv/2512.14967/) (2025-12-19): A new numerical method combines elicitability and deep learning for McKean-Vlasov stochastic equations, enabling efficient training of neural networks without expensive simulations, tested successfully on financial models. - [Adaptive Partitioning and Learning for Stochastic Control of Diffusion Processes](https://www.ml-quant.com/papers/arxiv/2512.14991/) (2025-12-19): An adaptive reinforcement learning algorithm enhances learning in controlled diffusion processes by partitioning state-action spaces, providing theoretical guarantees and effective results in applications like portfolio selection. - [Reinforcement Learning in Financial Decision Making: A Systematic Review of Performance, Challenges, and Implementation Strategies](https://www.ml-quant.com/papers/arxiv/2512.10913/) (2025-12-14): Reinforcement learning improves financial decision-making by simplifying complex investment problems, emphasizing clear explanations and strong reliability over complex algorithms. - [When Medical AI Explanations Help and When They Harm](https://www.ml-quant.com/papers/arxiv/2512.08424/) (2025-12-14): AIgenerated explanations can improve decision-making when algorithms are right, but can mislead when they're wrong, highlighting a paradox in AI transparency for doctors. - [Measuring Computer Science Enthusiasm: A Questionnaire-Based Analysis of Age and Gender Effects on Students' Interest](https://www.ml-quant.com/papers/arxiv/2512.08472/) (2025-12-14): Research indicates that age affects students' interest in computer science more than gender, suggesting educational approaches should consider developmental changes to boost engagement. - [The Adoption and Usage of AI Agents: Early Evidence from Perplexity](https://www.ml-quant.com/papers/arxiv/2512.07828/) (2025-12-14): A study of AI agent use with the Comet browser reveals that personal productivity and learning are the main reasons users interact with these tools. - [AI-Powered Direct Indexing: Exploring Thematic Universes for Enhanced Risk-Adjusted Returns](https://www.ml-quant.com/papers/ssrn/4977007/) (2025-12-01): The research presents FINDALL, a search engine that effectively identifies relevant stocks for direct indexing, outperforming traditional ETFs with lower costs. - [The Risk-Adjusted Intelligence Dividend: A Quantitative Framework for Measuring AI Return on Investment Integrating ISO 42001 and Regulatory Exposure](https://www.ml-quant.com/papers/arxiv/2511.21975/) (2025-12-01): Organizations using AI face challenges in traditional ROI calculations due to risks, leading to a new framework that includes risk assessments. - [Standardized Threat Taxonomy for AI Security, Governance, and Regulatory Compliance](https://www.ml-quant.com/papers/arxiv/2511.21901/) (2025-12-01): The use of AI in regulated industries exposes gaps between technical risk assessments and compliance, prompting a new taxonomy for AI risk evaluation and financial impacts. - [Tacit Bidder-Side Collusion: Artificial Intelligence in Dynamic Auctions](https://www.ml-quant.com/papers/arxiv/2511.21802/) (2025-12-01): Research shows that large language models can indirectly collaborate in Dutch auctions to increase prices, influenced by market structure. - [The Economics of AI Training Data: A Research Agenda](https://www.ml-quant.com/papers/arxiv/2510.24990/) (2025-11-04): Lays out data economics: why data is special, documents AI training-data deals, proposes a hierarchy of data units, and lists key research questions. - [Estimating Nationwide High-Dosage Tutoring Expenditures: A Predictive Model Approach](https://www.ml-quant.com/papers/arxiv/2510.24899/) (2025-11-04): Machine learning on incomplete ESSER plans estimates U.S. school districts spent about $2.2 billion on high‑dosage tutoring during COVID.