cs.AINov 17, 2026

The Only Edge Was Memorized in Training: LLM Trading Agents Around Earnings and Bad News

Authors: Jackson Wang

Abstract

Large language model (LLM) trading agents are tested on news, but news reaches stock prices within minutes, and a benchmark that replays headlines at their publication time finds no alpha in trading on them. Money can then remain in two places, before news whose date is known, such as earnings announcements, and after bad news, if the market's fall went too far. However, studies of LLMs and earnings read documents published with or after the announcement, and studies of reversal sort monthly portfolios, so neither tells whether an agent can trade these moments after costs. In this paper, we test LLM agents at both moments on the same look-ahead-free benchmark, before 598 earnings announcements and after 409 bad-news drops of large US stocks. Every input is cut at the decision close, staged agents are compared with single calls, simple rules and a logistic regression, and every decision is made with and without the stock's name. Across five models from four vendors, no agent beat the simple bets. Before earnings, the agents mostly bought and earned what buying before every announcement earned, and after bad news, a Gemini-based agent called almost every fall an overreaction while no agent picked better dips than buying them all. The only gains came from what the models had memorized in training. Given only the ticker and the dates, a GPT-based model called the direction of 59.2% of the announcements before its training cutoff and 53.3% after it. Agent builders should therefore compare an event agent with the simple bet and test it blind after the training cutoff.

Figures & tables

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

Nov 14, 2026

LLM Trading Agents Earn Beta, Not Alpha: A Look-Ahead-Free Benchmark

Large language model (LLM) trading agents now trade stocks for their users, and benchmarks rank models by their agents' returns. Yet a single run's return says little about skill. In our study, two runs of the same model that differed only in the random seed ended 5.2 points apart on average, so telling skill from luck takes many backtests across markets, models and seeds. However, live arenas trade each day only once, and backtests either use the few months after a model's training cutoff or run on years the model may remember, often with news joined to prices by calendar date. In this paper, we present rta-bench-1, a backtest over three US markets (a rally, a bear market and a crash). It shows each headline only after its publication minute, plays each period blind and named, and checks every run for look-ahead. It also splits each return into factor exposure and alpha and scores every decision against random choices of the same day and sector. In 91 runs of six models from four vendors, no alpha was significant, market and style exposure accounted for a median 78% of the return, and the decisions did no better than random stocks of the same sector. The agents bought the stocks with the most positive news, yet in ablations they earned as much without their headlines. Still, the agents carried out rules reliably, since a single cash rule in the prompt cut a GPT-based agent's share in stocks by 30.3 points in all six pairs. Builders should therefore supply the edge themselves and report exposures and alpha, and benchmarks should replay news by publication time.
May 27, 2026

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

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 end-to-end stock-market trading benchmark that addresses both issues. KTD-Fin uses a data-side masking protocol to anonymize key identifiers and calendar information consistently across prompts and tools, separating historical market memory from investment decision-making. It also incorporates a Barra-style performance attribution framework that decomposes portfolio returns into market, style, and stock-selection alpha components. Across ten frontier LLM agents evaluated on the Chinese CSI300 over a 2024--2026 window, masking substantially changes agent rationales, pushing them towards anonymized factor-based reasoning. Attribution analysis further shows that LLM agents' cumulative returns under leakage-controlled evaluation are largely explained by passive market and style exposure, with limited evidence of persistent stock-selection alpha. These findings suggest that financial LLM benchmarks should evaluate not only whether an agent makes money, but also whether the source of returns reflects transferable investment skill. We release KTD-Fin as a reproducible template for leakage-controlled and attribution-aware evaluation of LLM trading agents.
Nov 8, 2026

Inside the Trading Loop: Markets, Models, and Moves with LLM Agents

We built a hybrid multi-agent trading system that combines large language model (LLM) analysts with momentum-based portfolio construction and rule-based execution. Company-news and macroeconomic agents return structured reports to a coordinator that sets stock weights and cash reserves. We evaluate the portfolio engine using its broker-recorded paper account, which operated with deterministic fallbacks, and compare LLM-enabled configurations through retrospective replays. From March 11 to August 28, 2026, the account returned 34.5%, compared with 14.4% for the S&P 500 fund SPY over the same dates. This raw return advantage came with a larger maximum drawdown. Reconstructed policies also finish above matched SPY and QQQ references when assessments are used in the same session or deferred to the next. Design comparisons examine cash authority, selective invocation, and response contracts. A constant-cash control gives a return estimate close to stance-based sizing, and shorter report contracts improve observed parsing but can change recommendations.