Thinking Fast, Trading Late: System One Decision Models vs the Market

Ranked by how close their title and abstract are to this paper’s. Select some to read them together.

Closest papers

1. Very close Another approach to the problem Not relevant

   Trader-Company Method: A Metaheuristic for Interpretable Stock Price Prediction

   Katsuya Ito, Kentaro Minami, Kentaro Imajo et al. · 2020 · Trading and Market Microstructure · cited 16 times · AI summary available

   Trader-Company Method uses an interpretable metaheuristic to predict stock prices, while this paper tests language-model news signals under stricter evaluation controls.

   Abstract

   Investors try to predict returns of financial assets to make successful investment. Many quantitative analysts have used machine learning-based methods to find unknown profitable market rules from large amounts of market data. However, there are several challenges in financial markets hindering practical applications of machine learning-based models. First, in financial markets, there is no single model that can consistently make accurate prediction because traders in markets quickly adapt to newly available information. Instead, there are a number of ephemeral and partially correct models called "alpha factors". Second, since financial markets are highly uncertain, ensuring interpretability…

2. Very close Another approach to the problem Not relevant

   From Knowing to Doing: A Memory-Controlled Benchmark for LLM Trading Agents on Stock Markets

   Tao-Jie Zhu, Wen-Tao Zhao, Rui Sun et al. · 2026 · arXiv.org · preprint, not peer reviewed · cited 14 times · Not summarized yet

   KTD-Fin controls for memorized market knowledge and separates stock-selection skill from raw returns, addressing evaluation risks in LLM trading.

   Abstract

   Evaluating whether large language model (LLM) agents can profit in capital markets is increasingly framed as end-to-end trading: place an agent in a historical market, let it trade, and measure portfolio returns. This setup is vulnerable to two evaluation failures. First, long backtests often overlap with the knowledge cutoffs of frontier LLMs, allowing memorized tickers, dates, prices, and market narratives to substitute for investment reasoning. Second, raw returns are a noisy proxy for stock-selection ability, since positive performance may come from market beta, style exposure, or favorable regimes rather than genuine alpha. We introduce KTD-Fin (Knowing-To-Doing Financial Benchmark), an…

3. Very close Another approach to the problem Not relevant

   Learning to Trade Like an Expert: Cognitive Fine-Tuning for Stable Financial Reasoning in Language Models

   Yuchen Pan, S. Liew · 2026 · arXiv.org · preprint, not peer reviewed · Not summarized yet

   Cognitive Fine-Tuning trains LLMs with textbook and historical-market questions, whereas this paper tests news-based return signals with point-in-time controls.

   Abstract

   Recent deployments of large language models (LLMs) as autonomous trading agents raise questions about whether financial decision-making competence generalizes beyond specific market patterns and how it should be trained and evaluated in noisy markets lacking ground truth. We propose a structured framework for training and evaluating such models. Central to our approach is a curated, multiple-choice question (MCQ) dataset derived from classic textbooks and historical markets, verified by an AI committee, enriched with structured reasoning traces, and augmented to reduce shortcut learning. To evaluate whether performance on isolated MCQs generalizes to real-world trading, we introduce a two-st…

4. Close Broader context Not relevant

   Going All-In on LLM Accuracy: Fake Prediction Markets, Real Confidence Signals

   Michael Todasco · 2025 · arXiv.org · preprint, not peer reviewed · cited 2 times · Not summarized yet

   Fake Prediction Markets tests whether betting-game framing improves LLM forecasts of other models’ answers, rather than stock returns.

   Abstract

   Large language models are increasingly used to evaluate other models, yet these judgments typically lack any representation of confidence. This pilot study tests whether framing an evaluation task as a betting game (a fictional prediction market with its own LLM currency) improves forecasting accuracy and surfaces calibrated confidence signals. We generated 100 math and logic questions with verifiable answers. Six Baseline models (three current-generation, three prior-generation) answered all items. Three Predictor models then forecasted, for each question-baseline pair, if the baseline would answer correctly. Each predictor completed matched runs in two conditions: Control (simple correct/i…

5. Close Another approach to the problem Not relevant

   StockBench: Can LLM Agents Trade Stocks Profitably In Real-world Markets?

   Yanxu Chen, Zijun Yao, Yantao Liu et al. · 2025 · Machine Learning · preprint, not peer reviewed · cited 49 times · AI summary available

   StockBench evaluates LLM agents in realistic iterative trading, while this paper tests whether news-based signals predict returns and remain actionable at the open.

   Abstract

   Large language models (LLMs) have recently demonstrated strong capabilities as autonomous agents, showing promise in reasoning, tool use, and sequential decision-making. While prior benchmarks have evaluated LLM agents in domains such as software engineering and scientific discovery, the finance domain remains underexplored, despite its direct relevance to economic value and high-stakes decision-making. Existing financial benchmarks primarily test static knowledge through question answering, but they fall short of capturing the dynamic and iterative nature of trading. To address this gap, we introduce StockBench, a contamination-free benchmark designed to evaluate LLM agents in realistic, mu…

6. Close Same question Not relevant

   Can LLM-based Financial Investing Strategies Outperform the Market in Long Run?

   Weixian Waylon Li, Hyeonjun Kim, Mihai Cucuringu et al. · 2025 · Knowledge Discovery and Data Mining · cited 32 times · Not summarized yet

   FINSABER tests whether LLM investing strategies outperform across longer periods and broader stock universes, examining robustness to biases in backtests.

   Abstract

   Large Language Models (LLMs) have recently been leveraged for asset pricing and stock trading applications, enabling AI agents to generate investment decisions from unstructured financial data. However, most evaluations of LLM timing-based investing strategies are conducted on narrow timeframes and limited stock universes, overstating effectiveness due to survivorship and data-snooping biases. We critically assess their generalizability and robustness by proposing FINSABER, a backtesting framework evaluating timing-based strategies across longer periods and a larger universe of symbols. Systematic backtests over two decades and 100+ symbols reveal that previously reported LLM advantages dete…

7. Close Another approach to the problem Not relevant

   AlphaForgeBench: Benchmarking End-to-End Trading Strategy Design with Large Language Models

   Wen-Tao Zhang, Ming-Xuan Zhao, Jin-Cheng Gao et al. · 2026 · Knowledge Discovery and Data Mining · cited 6 times · Not summarized yet

   AlphaForgeBench evaluates end-to-end strategy design and trading-agent stability, rather than point-in-time prediction from masked company news.

   Abstract

   The rapid advancement of Large Language Models (LLMs) has catalyzed the proliferation of diverse financial benchmarks, progressively evolving from static knowledge evaluation to increasingly sophisticated interactive trading simulations. Nevertheless, existing frameworks that assess real-time trading performance largely overlook a fundamental failure mode: the severe behavioral instability exhibited by LLMs in sequential decision-making under financial uncertainty. Through extensive empirical investigation, we demonstrate that when deployed as direct trading agents, LLMs manifest extreme run-to-run variance, produce inconsistent action sequences even under strictly deterministic decoding con…

8. Close Broader context Not relevant

   Reflexivity as Prompt: Does Awareness of Self-Reinforcing Market Dynamics Improve LLMs as Financial Market Forecasters?

   Eugene W Park · 2026 · Not summarized yet

   Reflexivity as Prompt studies whether explaining self-reinforcing market dynamics improves LLM forecasts across historical boom-bust episodes.

   Abstract

   We study how frontier large language models (LLMs) behave as financial forecasters during boom-bust market cycles when made progressively aware of Soros's theory of reflexivity. Standard AI-assisted forecasting treats the market as an exogenous system. Reflexivity theory holds otherwise: prices shape fundamentals, and every forecaster is a participative agent in the loop it analyzes. We evaluate three frontier models - GPT5, Claude Sonnet 4.6, and Gemini 3 Pro - under four accumulating zero-shot conditions across two historically distinct episodes: the dot-com bubble (1996-2001) and the global financial crisis (2004-2009). The primary metric is directional forecasting accuracy; we also repor…

Prior work it builds on

1. Close Same question Not relevant

   Pervasive underreaction: Evidence from high-frequency data

   Hao Jiang, Sophia Zhengzi Li, Hao Wang · 2020 · Journal of Financial Economics · cited 42 times · Not summarized yet

   Pervasive underreaction tests whether stock prices drift after firm news, while this paper finds news priced by the open and no reliable evidence of underreaction.

   Abstract

   We propose a novel high-frequency decomposition of daily stock returns into news- and non-news-driven components. With it, we uncover evidence of pervasive stock market underreaction to firm news, in the post-millennium period when stocks appear to be priced with increased efficiency. We show that stock prices tend to drift in the same direction as the initial market response, for several days after the news arrival without reversals. A trading strategy exploiting the return drift generates a monthly return of 2.10% after trading costs. To understand the economic mechanism, we find that the return drift is stronger when investors are distracted. Using analyst earnings forecasts as a proxy, w…

2. Close Same method, another problem Not relevant

   Approaching Human-Level Forecasting with Language Models

   Danny Halawi, Fred Zhang, Chen Yueh-Han et al. · 2024 · Machine Learning · cited 165 times · AI summary available

   Approaching Human-Level Forecasting uses retrieval-augmented language models to generate forecasts, applying language-model prediction to future events rather than stock news.

   Abstract

   Forecasting future events is important for policy and decision making. In this work, we study whether language models (LMs) can forecast at the level of competitive human forecasters. Towards this goal, we develop a retrieval-augmented LM system designed to automatically search for relevant information, generate forecasts, and aggregate predictions. To facilitate our study, we collect a large dataset of questions from competitive forecasting platforms. Under a test set published after the knowledge cut-offs of our LMs, we evaluate the end-to-end performance of our system against the aggregates of human forecasts. On average, the system nears the crowd aggregate of competitive forecasters, an…

3. Close Same method, another problem Not relevant

   Wisdom of the Silicon Crowd: LLM Ensemble Prediction Capabilities Match Human Crowd Accuracy

   Philipp Schoenegger, Indre Tuminauskaite, Peter S. Park et al. · 2024 · Computers and Society · cited 131 times · AI summary available

   Wisdom of the Silicon Crowd aggregates LLM predictions to forecast events, rather than testing individual model signals on stock returns.

   Abstract

   Human forecasting accuracy in practice relies on the 'wisdom of the crowd' effect, in which predictions about future events are significantly improved by aggregating across a crowd of individual forecasters. Past work on the forecasting ability of large language models (LLMs) suggests that frontier LLMs, as individual forecasters, underperform compared to the gold standard of a human crowd forecasting tournament aggregate. In Study 1, we expand this research by using an LLM ensemble approach consisting of a crowd of twelve LLMs. We compare the aggregated LLM predictions on 31 binary questions to that of a crowd of 925 human forecasters from a three-month forecasting tournament. Our main anal…

4. Close Groundwork for this paper Not relevant

   The Memorization Problem: Can We Trust LLMs'Economic Forecasts?

   Alejandro Lopez-Lira, Yue-Hua Tang, Ming Zhu · 2025 · cited 50 times · Not summarized yet

   The Memorization Problem shows that LLMs recall financial data and reconstruct masked entities, motivating this paper’s masking and look-ahead controls.

   Abstract

   Large language models (LLMs) cannot be trusted for economic forecasts during periods covered by their training data. Counterfactual forecasting ability is non-identified when the model has seen the realized values: any observed output is consistent with both genuine skill and memorization. Any evidence of memorization represents only a lower bound on encoded knowledge. We demonstrate LLMs have memorized economic and financial data, recalling exact values before their knowledge cutoff. Instructions to respect historical boundaries fail to prevent recall-level accuracy, and masking fails as LLMs reconstruct entities and dates from minimal context. Post-cutoff, we observe no recall. Memorizatio…

5. Close Groundwork for this paper Not relevant

   Information, Trading, and Volatility: Evidence from Firm-Specific News

   J. Boudoukh, Ronen Feldman, Shimon Kogan et al. · 2018 · The Review of financial studies · cited 152 times · Not summarized yet

   Information, Trading, and Volatility measures firm-specific news information in stock returns, providing a basis for this paper’s study of news-based signals.

   Abstract

   What moves stock prices? Prior literature concludes that the revelation of private information through trading, and not public news, is the primary driver. We revisit the question by using textual analysis to identify fundamental information in news. We find that this information accounts for 49.6% of overnight idiosyncratic volatility (vs. 12.4% during trading hours), with a considerable fraction due to days with multiple news types. We use our measure of public information arrival to reinvestigate two important contributions in the literature related to individual R2s of stock returns on aggregate factors. (JEL G14)