{"paper_id":"2504.10789","title":"Can Large Language Models Trade? Testing Financial Theories with LLM Agents in Market Simulations","authors":["Alejandro Lopez-Lira"],"paper_date":"2025-04-15","paper_type":"empirical","paper_url":"https://arxiv.org/pdf/2504.10789v1","page_url":"https://thequant.space/flowcharts/can-large-language-models-trade-testing-financial-theories/","scores":{"empirical_rigor":8.5,"math_complexity":4,"quadrant":"Street Traders"},"proposes":{"status":"available","statement":"The paper introduces an open‑source simulated stock market framework that lets large language models act as heterogeneous trading agents with a persistent order book and realistic market mechanisms.","quote":"This paper presents a realistic simulated stock market where large language models (LLMs) act as heterogeneous competing trading agents. The open-source framework incorporates a persistent order book with market and limit orders, partial fills, dividends, and equilibrium clearing alongside agents with varied strategies, information sets, and endowments.","summary":"This paper presents a realistic simulated stock market where large language models (LLMs) act as heterogeneous competing trading agents. The open-source framework incorporates a persistent order book with market and limit orders, partial fills, dividends, and equilibrium clearing alongside agents…","strategy":null},"assumptions":{"status":"available","items":[{"assumption":"The market operates as a continuous double‑auction processed in discrete trading rounds.","type":"market","stated":"explicit","quote":"2.1.1 Market Mechanism Design Our framework implements a continuous double-auction market mechanism that processes orders in discrete trading rounds. We use discrete trading rounds because LLMs have latency constraints, making it infeasible to process orders in real time.","support":"full"},{"assumption":"Agent order submission order is randomized each round to avoid systematic priority.","type":"execution","stated":"explicit","quote":"Within each round, the order of agent submissions is randomized to avoid giving systematic priority to specific agents, thus simulating concurrent order arrival while maintaining fairness.","support":"full"},{"assumption":"Short selling is not allowed for agents.","type":"execution","stated":"explicit","quote":".2.14 Trading Options Trading Options: New Orders (replace_decision=’Add’): Single or multiple orders allowed For each order: Market order: Set order_type=’market’ Limit order: Set order_type=’limit’ and specify price_limit IMPORTANT: Sell orders require sufficient available shares Short selling is NOT allowed Cancel Orders (replace_decision=’Cancel’): Return an empty orders list: orders=[] The trading options","support":"full"},{"assumption":"Cash holdings earn a risk‑free interest rate of 5% per round, providing an opportunity cost for asset investment.","type":"model","stated":"explicit","quote":"Cash holdings earn a risk-free interest rate of 5% per round, providing an opportunity cost for asset investment.","support":"full"},{"assumption":"In finite‑horizon markets agents know the total number of rounds and redeem shares at the fundamental value at the final round.","type":"market","stated":"explicit","quote":"In finite-horizon markets, agents are informed of the total number of rounds, and their terminal wealth is calculated by redeeming all shares at the fundamental value of the final round. In infinite-horizon markets, no terminal information is provided to agents, and final wealth is determined using the last market price for share valuation.","support":"full"},{"assumption":"Fundamental value follows a dividend discount model with expected dividend $1.40 and interest rate 5%, yielding a constant fundamental value of $28.00.","type":"model","stated":"explicit","quote":"markets, the fundamental value follows the standard dividend discount model: Vt = E[D] r = Expected Dividend Interest Rate (1) For the default parameters where the expected dividend is $1.40 and the interest rate is 5%, this yields a fundamental value of $28.00. In finite horizon markets with T periods remaining, the fundamental value incorporates 30 both future dividends and terminal redemption value K: FVt = T X","support":"full"}]},"datasets":{"status":"available","items":[{"name":"Synthetic market simulation data","provider":null,"kind":"synthetic","access":"unknown","period":null,"universe":null,"quote":"markets, the fundamental value follows the standard dividend discount model: Vt = E[D] r = Expected Dividend Interest Rate (1) For the default parameters where the expected dividend is $1.40 and the interest rate is 5%, this yields a fundamental value of $28.00. Each scenario maintains the fundamental value of $28.00 while varying starting prices, time horizons, and agent populations.","support":"full"}],"known":[],"sample":{"start":null,"end":null,"market":null,"asset_class":null,"frequency":null},"data_resolution":"synthetic"},"transaction_costs":{"status":"none_found","modeled":"unclear","headline_net_of_costs":null,"detail":null,"quote":null,"source":"none","rule":"R0b","candidates":{"front_matter":"unclear","evidence_card":"no","evidence_card_quoted":false},"meaning":"no cost treatment was found in the passages the extractor read; that is not evidence that costs are not modeled"},"results":{"status":"available","evaluation_type":"none","findings":[{"finding":"LLMs demonstrate consistent strategy adherence and can act as value investors, momentum traders, or market makers as instructed.","kind":"performance","quote":"Three findings emerge: First, LLMs demonstrate consistent strategy adherence and can function as value investors, momentum traders, or market makers per their instructions.","support":"full"},{"finding":"Market dynamics in the simulations exhibit real‑world features such as price discovery, bubbles, underreaction, and strategic liquidity provision.","kind":"other","quote":"Second, market dynamics exhibit features of real financial markets, including price discovery, bubbles, underreaction, and strategic liquidity provision.","support":"full"},{"finding":"The framework enables systematic analysis of LLM responses to varying market conditions, akin to partial dependence plots.","kind":"theoretical","quote":"Third, the framework enables analysis of LLMs' responses to varying market conditions, similar to partial dependence plots in machine-learning interpretability.","support":"full"},{"finding":"LLMs can effectively execute trading strategies, understanding market mechanics, forming price expectations, and following prompts regardless of profit implications.","kind":"performance","quote":"First, LLMs can effectively execute trading strategies. They consistently understand market me-chanics, process market information, form price expectations, and execute trades according to specific instructions.","support":"full"},{"finding":"Interactions among LLM agents generate realistic market dynamics, including price bubbles and corrections, highlighting both capabilities and potential systemic risks.","kind":"other","quote":"Our experimental framework reveals that their in-teractions generate realistic market dynamics, including phenomena like price bubbles and corrections, highlighting both the capabilities and potential risks of deploying LLM agents in financial markets. While LLM agents can enhance price discovery and liquidity, their adherence to programmed strate-gies, even potentially flawed ones derived from prompts, could","support":"full"}],"reported_performance":null,"out_of_sample":{"method":"none","headline_out_of_sample":null}},"independent_evidence":{"status":"available","supports":[],"contradicts":[{"paper":{"paper_id":"2502.15800","title":"LLM Agents Do Not Replicate Human Market Traders: Evidence From Experimental Finance","paper_date":"2025-02-18","authors":["Thomas Henning","Siddhartha M. Ojha","Ross Spoon","Jiatong Han","Colin F. Camerer"],"empirical_rigor":7,"page_url":"https://thequant.space/flowcharts/llm-agents-do-not-replicate-human-market-traders-evidence-f/"},"question":"Do LLM agents replicate human market trader behavior in asset pricing and bubble formation?","this_paper_finding":"Interactions among LLM agents generate realistic market dynamics, including price bubbles and corrections, highlighting both capabilities and potential systemic risks.","other_paper_finding":"LLM agents generally price the asset near its fundamental value, showing “textbook‑rational” behavior and only muted bubble formation.","other_paper_quote":"LLM‑agents typically converge on prices near the fundamental value, thereby exhibiting more “textbook-rational” behavior than human subjects.","note":"A shows near‑fundamental pricing with muted bubbles; B shows realistic dynamics with bubbles and corrections, indicating a contradiction.","confidence":0.494},{"paper":{"paper_id":"2510.01451","title":"Financial Stability Implications of Generative AI: Taming the Animal Spirits","paper_date":"2025-10-01","authors":["Anne Lundgaard Hansen","Seung Jung Lee"],"empirical_rigor":6.5,"page_url":"https://thequant.space/flowcharts/financial-stability-implications-of-generative-ai-taming-the-animal-spirits/"},"question":"Do AI/LLM agents mitigate or amplify asset‑price bubbles in financial markets?","this_paper_finding":"Interactions among LLM agents generate realistic market dynamics, including price bubbles and corrections, highlighting both capabilities and potential systemic risks.","other_paper_finding":"AI agents exhibit less herd behavior than human financial professionals, suggesting a potential reduction in asset‑price bubbles.","other_paper_quote":"Overall, our results suggest that AI agents exhibit less herd behavior compared to human financial professionals. First, less herding behavior could lead to fewer extreme market movements and asset price bubbles.","note":"A shows LLM agents create realistic bubbles and corrections, while B finds AI agents herd less, implying reduced bubbles—contradictory findings.","confidence":0.461}],"qualifies":[],"compared_with":29,"citations":{"cited_by_in_corpus":6,"by_independent_authors":6,"how_cited":{"used_as_baseline":0,"uses_method_or_data":0,"extends":0,"discusses_findings":0,"mentions":5,"unclassified":1},"independent_papers":[],"scope":"Citations between papers in this corpus only; citations from outside it are not counted."}},"implementation":{"status":"available","public":false,"code_url":null,"note":"No repository presented by the authors as their own was found in the paper. Absence here is not proof that no code exists."},"replication":{"status":"none_found","card":null,"note":"No replication of this paper has been run by thequant.space yet."},"extraction":{"text_source":"fulltext","model":"openai/gpt-oss-120b","version":2,"extracted_at":"2026-11-14T19:50:46Z"},"provenance":{"claims":"Everything here is what the authors state in the paper. Nothing was re-run or reproduced by the platform.","quotes":"Assumptions, datasets, cost detail and findings were extracted by a language model. Every quote is verbatim paper text. Where an item has `support`, \"full\" means every figure in it appears in the quoted sentences and \"partial\" means a figure is stated elsewhere in the paper; items with a figure found nowhere in the paper were discarded.","datasets_known":"datasets.known lists well-known vendors and public datasets by a stable id, found by a dictionary over the full text where the sentence says the data is used; it is independent of the model and covers only names in the dictionary.","flags":"transaction_costs.modeled, results.evaluation_type and results.out_of_sample.method are platform classifications (structured-decision models); see https://thequant.space/score-guide/#penalty-flags","comparisons":"supports / contradicts / qualifies are a language model's reading of two papers' quoted findings, restricted to closely related papers that share no author. They are pointers to read, not verified replications, and coverage is partial.","implementation":"code_url is a repository the paper itself presents as its own (rule-scored, ambiguous cases judged by a model), checked for liveness when it was recorded."},"docs":"https://thequant.space/agents/"}