Financial language models improve news-driven stock trading under real market conditions
Financial Language Models as Applied Artificial Intelligence Systems for News-Based Trading under Market Frictions
Computation and LanguageMachine Learning
Summary
Translating financial news into useful signals for stock trading can be hard because of timing, costs, and market limits. The authors created a framework called MFAST that connects financial news to stock price data, tests models only on new news after a certain date, and checks how well these models predict stock returns while accounting for real trading challenges. They found that certain language models work better than others at turning news into trading decisions that make money. Their work shows the importance of testing trading models with real-world factors from start to finish.
What this means in practice
- •For quantitative traders: Use MFAST to build and validate news-based trading strategies that account for real trading costs and market constraints improving portfolio returns.
- •For financial data engineers: Integrate timestamped news feeds and price data using the MFAST framework to create reproducible, auditable financial AI pipelines for trading decisions.
- •For financial software developers: Develop AI-powered market analysis tools that convert financial text sentiment into actionable trading signals with operational diagnostics on cost and latency.$Commercial implications: Enables building commercial trading platforms with validated AI components optimized for market conditions.
Authors
Kemal Kirtac
Abstract
Financial language models can transform unstructured firm-specific news into structured decision signals, but financial AI research lacks an integrated deployment framework for evaluating whether those signals remain useful in financial decision systems. Computer science research has developed strong methods for time-series forecasting, text classification, multimodal stock prediction, graph-based market modeling, and machine-learning operations, yet these streams do not provide a domain-specific protocol that jointly tests financial language-model outputs under event-time observability, probability calibration, execution timing, transaction costs, liquidity constraints, capacity limits, operational diagnostics, and statistical inference. We introduce MFAST, a Market-Friction-Aware Sentiment-to-Trading framework that converts timestamped financial text into auditable, reproducible, and market-feasible trading decisions. The application is news-based trading, where firm-specific text must be linked to securities before portfolio decisions can be evaluated. The framework links Refinitiv News Analytics to Center for Research in Security Prices (CRSP) equity data, restricts the primary out-of-sample evaluation to post-release news outside disclosed foundation-model data-freshness periods, and adds a public replication arm using open financial text and public price data. Results show that decoder-only language models outperform encoder baselines and dictionary sentiment in classification, calibration, return prediction, and net portfolio performance, while operational diagnostics reveal trade-offs among accuracy, latency, memory, throughput, and inference cost. The paper shows that credible evaluation of financial language models requires an end-to-end engineering approach combining language understanding, temporal discipline, market-friction-aware deployment, and reproducible validation.