RetailAgent: Structured Adverse Timing in Self-Conditioned Multimodal LLM Trading Agents

Yupeng Zhang; Liuyuan Jiang; Hongyi Huang; Bingheng Li; Lisha Chen

cs.AIAug 28, 2026

RetailAgent: Structured Adverse Timing in Self-Conditioned Multimodal LLM Trading Agents

Authors: Yupeng Zhang, Liuyuan Jiang, Hongyi Huang, Bingheng Li, Lisha Chen

Organizations: University of Wisconsin–Madison · University of Rochester · Michigan State University

Abstract

In financial markets, a sequential policy that reacts systematically to price movements may become predictable to other market participants. This paper studies whether large language model (LLM) agents exhibit such directional structure through RetailAgent, an experimental framework in which an LLM observes anonymized intraday equity price histories and permitted state, then repeatedly chooses long (hold the stock) or flat (stay out) before the subsequent interval return is revealed. We compare returns during long and flat intervals along the same stock's intraday path after removing the overall fraction of long decisions. This exposure-matched measure reveals persistent negative timing across modality, horizon, state, and model family. Shuffling saved action sequences substantially attenuates the effect, showing that alignment between actions and subsequent returns drives the negative score. Feeding self-authored memories into decisions further increases policy persistence, while timing becomes more negative among stock-days on which the agent uses both actions. These results reveal stable, recoverable directional structure in sequential LLM financial decisions and a behavioral signal for studying how another participant could respond to a predictable policy.

Figures & tables

Figure 1: Top: price-conditioned behavior may invite a counter-response in an interactive market. Bottom: RetailAgent isolates a policy, chooses a binary action from permitted state, and is evaluated on revealed returns.

Model or control	Observation or state	Interval	Stock-days	Timing alpha, bps (95% CI)	Switches
A. Fourteen standard configurations
Qwen3.5	price text	10m	13,710	$-45.7\,[-49.4,-41.8]$
Qwen3.5	chart	10m	14,937	$-29.9\,[-33.7,-26.2]$
Qwen3.5	multimodal	10m	14,438	$-48.9\,[-52.7,-45.0]$
Qwen3.5	price text	20m	23,594	$-56.7\,[-59.8,-53.6]$
Qwen3.5	+position	20m	25,410	$-54.7\,[-58.2,-51.0]$

Table 1: Exposure-matched timing alpha. Principal local text, chart, and multimodal rows use Qwen/Qwen3.5-9B ( Qwen Team, 2026 ) . The coarse-image archive is recorded as Qwen3-VL ( Bai et al., 2025 ) . Other models are Claude Haiku 4.5 ( Anthropic, 2025 ) and Granite-4.0-H-Small ( IBM Research, 2025 ) . Confidence intervals resample dates. Switches count long/flat changes per stock-day. Granite rows are archived point estimates. Cross-frame magnitudes are descriptive.

Comparator	Stock-days	Corr. with $q$	Inverted	Residual timing (95% CI)
GRU price model	2,519	$+0.114$	$+46.8$	$+43.9\,[38.1,49.5]$
GBDT-38	13,717	$+0.132$	$+45.7$	$+40.4\,[37.0,43.8]$
One-period reversal	13,717	$+0.286$	$+45.7$	$+38.9\,[36.0,41.8]$

Table 2: Inverted text-agent timing after within-stock projection on conventional price signals.

Self-conditioning condition	Scored $N$	Timing alpha (95% CI)	IC (95% CI)
No directive, $w=5$	1,289	$-59.9\,[-67.1,-53.0]$	$-0.0105\,[-0.0233,0.0017]$
Conviction, $w=5$	1,011	$-58.1\,[-67.6,-48.6]$	$+0.0028\,[-0.0082,0.0128]$
Consistency, current only ( $w=0$ )	1,194	$-62.8\,[-69.7,-56.0]$	$-0.0134\,[-0.0278,0.0009]$
Consistency, one earlier ( $w=1$ )	1,054	$-74.1\,[-81.5,-66.6]$	$-0.0125\,[-0.0260,0.0011]$
Consistency, $w=5$	1,154	$-66.7\,[-73.9,-59.7]$	$-0.0087\,[-0.0212,0.0036]$
Consistency, $w=12$	1,184	$-67.1\,[-75.0,-59.6]$	$-0.0076\,[-0.0190,0.0039]$

Table 3: Narrative self-conditioning results. Every main-period condition attempts 1,500 stock-days across 150 dates. Scored

N

excludes failed intervals and constant action trajectories. The upper panel reports conditional timing alpha in bps per stock-day and contemporaneous cross-sectional IC with 95% date-clustered confidence intervals. The lower panel reports later-period point estimates and observed long/flat position changes.

	Conventional		Self-conditioning
Metric pair	Pearson	Spearman	Pearson	Spearman
Timing vs portfolio return	$+0.355$	$+0.392$	$-0.066$	$+0.018$
Timing vs cross-sectional IC	$+0.684$	$+0.399$	$-0.075$	$-0.212$
Portfolio return vs cross-sectional IC	$+0.859$	$+0.881$	$+0.920$	$+0.867$

Table 4: Condition-level metric correlations. Conventional contains 12 standard conditions. Self-conditioning contains ten memory conditions. Each condition contributes one observation.

Shared centered-outcome group
Configuration	Timing score, bps (95% CI)
Soft tokens disabled	$-21.0\,[-23.8,-18.3]$
Two tokens, 92-word basis	$-33.7\,[-37.5,-30.0]$
Eight tokens, 92-word basis	$-32.3$
Sixteen tokens, 92-word basis	$-29.2$
Two tokens, 492-word basis	$-33.8\,[-37.5,-30.1]$

Table 5: Latent-state ablations. Comparisons within the upper group share a training outcome. Lower-group rows use separate transformed objectives or evaluation frames. Date-clustered 95% confidence intervals appear where available.

Condition	Long intervals	Exposure	Long return	Flat return	Timing alpha
Price text	$7/23$	$0.304$	$-576.7$	$+575.9$	$-576.4$
Consistency, one earlier memory	$16/23$	$0.696$	$+359.4$	$+1{,}404.0$	$-867.3$

Table 6: Trajectory examples. Returns and timing alpha are in bps under the research-return labels. The rows illustrate the estimand on stock-days and support the evidence in Tables 1 and 3 .

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Narrative condition	Failed intervals	Affected cells	Constant paths	Scored $N$
No directive, $w=5$	1,120	641	211	1,289
Conviction, $w=5$	1,690	584	489	1,011
Consistency, $w=0$	1,581	925	306	1,194
Consistency, $w=1$	1,545	615	446	1,054
Consistency, $w=5$	505	344	346	1,154
Consistency, $w=12$	266	207	316	1,184

Appendix

Table A1: Narrative scoring provenance. Affected cells contain at least one failed interval. Constant paths remain all long or all flat after failed intervals are removed. The archived conditional timing estimates use the scored stock-days in the final column.

Figure B1: Standard and representative self-conditioning timing estimates for the locally cached Qwen/Qwen3.5-9B checkpoint and the archived Qwen3-VL coarse-image condition [ Qwen Team, 2026 , Bai et al., 2025 ] , with the Claude Haiku 4.5 sign check [ Anthropic, 2025 ] . Points report within-stock timing alpha with 95% date-clustered confidence intervals. Table 1 reports sample sizes and exact values.

Figure B2: Within-stock timing alpha and contemporaneous cross-sectional IC for ten Qwen3.5-9B narrative self-conditioning conditions [ Qwen Team, 2026 ] . Each condition attempts 1,500 stock-days. Scored counts appear in Table 3 . Both panels report 95% date-clustered confidence intervals.

Algorithmic Trading

Algorithmic trading uses computational models to generate signals, allocate portfolios, and execute orders across financial markets.

See all 151 papers

Agentic Trading: When LLM Agents Meet Financial Markets
May 19, 2026Yihan Xia, Panpan You, Taotao Wang +4LLM Agent Evaluation Finance
College of Electronic and Information Engineering, Shenzhen University, Shenzhen, China
Representation Signatures and Risk-Feedback Alignment in LLM Trading Agents
May 16, 2026Weicheng XueLLM Alignment LLM Auditing
Virginia Tech
FinPersona-Bench: A Benchmark for Longitudinal Psychometric Stability of Autonomous Financial Agents
Jun 30, 2026Muhammad Usman Safder, Ayesha Gull, Rania Elbadry +7Personality Modeling in Language Models Long-Horizon Agent Evaluation
Steve · MBZUAI · The University of Tokyo +2

May 19, 2026J/K move · Enter open · S save

Agentic Trading: When LLM Agents Meet Financial Markets

Yihan Xia, Panpan You, Taotao Wang, Fang Liu+3

College of Electronic and Information Engineering, Shenzhen University, Shenzhen, China

LLM Agent Evaluation Finance Algorithmic Trading

Open paper AskPDF arXiv

A growing body of work explores how Large Language Models (LLMs) can be embedded in trading systems as agents that perceive market information, retrieve context, reason about decisions, emit tradable actions, and adapt under market feedback. This paper reframes LLM-based trading agents as expert-system decision pipelines and presents an audit-oriented evidence map of 77 included studies in a protocol-coded snapshot screened through 2026-03-09. A primary empirical subset (n=19) satisfies the minimum boundary of Action Output plus Closed-Loop Evaluation; the remaining 58 included studies are retained as background and design context. The central empirical finding is protocol incomparability: within the primary subset, only 2/19 studies report extractable time-consistent split protocols, 1/19 reports an explicit transaction-cost model, 1/19 documents universe or survivorship handling, 11/19 report execution timing or semantics, 15/19 are coded as R0, and no study reaches R3 reproducibility. We therefore use Architecture-Capability-Adaptation as a working analytical lens rather than a validated taxonomy, and we foreground the evidence ledger, reproducibility audit, and reporting checklist as the main contributions. The resulting survey shows that architectural experimentation is expanding rapidly, while comparable evaluation protocols, execution semantics, and reproducible artifacts remain the field's immediate bottlenecks.

Abstract

Algorithmic Trading

Explore similar work

Agentic Trading: When LLM Agents Meet Financial Markets

Representation Signatures and Risk-Feedback Alignment in LLM Trading Agents

FinPersona-Bench: A Benchmark for Longitudinal Psychometric Stability of Autonomous Financial Agents