Learning AI Agents, Market Makers, and Stress in Automated Markets

Irene Aldridge

econ.EMNov 1, 2026

Learning AI Agents, Market Makers, and Stress in Automated Markets

Authors: Irene Aldridge

Organizations: RiskAICenter and Cornell Tech, Cambridge and Johns Hopkins

Abstract

Automated markets increasingly route order flow from learning trading agents through automated market makers (AMMs). We model trader archetypes as AI agents defined by their reward functions and analyze the feedback they create on a constant-product AMM. Herd-dominated and arbitrage-dominated markets of equal size have identical return distributions and turnover, so no rule based on volatility, value-at-risk, stress frequency, or volume can tell them apart; yet their stress episodes differ in length by a factor unbounded in the feedback-to-noise ratio $β/s$ . Serial statistics identify the feedback sign from $Θ((s/β)^2\log(1/δ))$ observations, and this rate is tight. When agents trade together, profit-seeking agents ride herds rather than lean against them, leading to all-or-nothing participation. Spread-earning market makers restore stability but make liquidity fragile, with discontinuous collapse and hysteresis; maker--taker pricing, not a uniform fee, is the effective instrument.

Figures & tables

Figure 1: (a) Theorem 4.2 : expected stress-episode length (

c=0

, Gaussian news) in a herd-dominated and an arbitrage-dominated market. The two have identical volatility and stress frequency at every

|\beta|

. (b) Theorem 6.1 (v): expected down-run length against rider capital

\bar{X}

(

\lambda=s=1

h=0.3

\phi=0.2

\bar{\sigma}=1

). Market makers exit at

\bar{X}_{c}=0.51

and return only below

\bar{X}_{lo}=0.21

; in the shaded band both outcomes are stable.

Appendix figures & tables1 asset

Supplementary material from the paper’s appendix.

Appendix

Result and setting	Simulated (theory)
Thm. 4.2 , $\beta=+0.8$ , $c=1$ , $2\times 10^{6}$ periods	$\operatorname{Var}(r)$ $1.6398$ ( $1.6400$ ); $F$ $0.2282$ ( $0.2283$ ); $\mathbb{E}[L]$ $1.7247$ ( $1.7263$ ); $q$ $0.7883$ ( $0.7881$ )
Thm. 4.2 , $\beta=-0.8$ , $c=1$	$\operatorname{Var}(r)$ $1.6398$ ( $1.6400$ ); $F$ $0.2287$ ( $0.2283$ ); $\mathbb{E}[L]$ $1.0372$ ( $1.0373$ ); $q$ $0.2117$ ( $0.2119$ )
Prop. 4.6 , $\Delta=\tfrac{1}{2}$ , $3\times 10^{5}$ periods	longest own-stress episode $2$ ( $\leq 2$ ); mean $1.08$
Lemma 5.1 , $h=0.3$ , $(x_{M},x_{C})=(0,0)$ , $10^{6}$ periods	rider gross $0.1847$ ( $0.1854$ ); $q$ $0.6181$ ( $0.6179$ ); $\mathbb{E}[L]$ $2.6164$ ( $2.6172$ )
$(x_{M},x_{C})=(0.5,0)$	$0.2102$ ( $0.2108$ ); $0.7026$ ( $0.7027$ ); $3.3608$ ( $3.3638$ )
$(x_{M},x_{C})=(1,0)$	$0.2155$ ( $0.2164$ ); $0.7215$ ( $0.7214$ ); $3.5860$ ( $3.5891$ )

Appendix

Table 1: Simulated versus theoretical values. Payoffs are gross of fees, per unit per period.

Market Making

See all 9 papers

MarketBench: Evaluating AI Agents as Market Participants
Apr 26, 2026Andrey Fradkin, Rohit KrishnanMulti-Agent Coordination LLM Agent Evaluation
Boston University and the MIT Initiative on the Digital Economy. · Independent Researcher.
Dissecting AI Trading: Behavioral Finance and Market Bubbles
Apr 20, 2026Shumiao Ouyang, Pengfei SuiQuantitative Finance Algorithmic Trading
Saïd Business School, University of Oxford · School of Management and Economics, The Chinese University of Hong Kong, Shenzhen
Information Aggregation with AI Agents
Apr 21, 2026Spyros GalanisAI Agent Evaluation Multi-Agent Reasoning
Department of Economics, University of Durham

Apr 26, 2026J/K move · Enter open · S save

MarketBench: Evaluating AI Agents as Market Participants

Andrey Fradkin, Rohit Krishnan

Boston University and the MIT Initiative on the Digital Economy. · Independent Researcher.

Multi-Agent Coordination LLM Agent Evaluation AI Agent Benchmarks

Open paper AskPDF arXiv

Markets are a promising way to coordinate AI agent activity for similar reasons to those used to justify markets more broadly. In order to effectively participate in markets, agents need to have informative signals of their own ability to successfully complete a task and the cost of doing so. We propose MarketBench, a benchmark for assessing whether AI agents have these capabilities. We use a 93-task subset of SWE-bench Lite, a software engineering benchmark, with six recently released LLMs as a demonstration. These LLMs are miscalibrated on both success probability and token usage, and auctions built from these self-reports diverge from a full-information allocation. A follow-up intervention where we add information about capabilities from prior experiments to the context improves calibration, but only modestly narrows the gap to a full-information benchmark. We also document the performance of a market-based scaffolding with these LLMs. Our results point to self-assessment as a key bottleneck for market-style coordination of AI agents.

Abstract

Market Making

Explore similar work

MarketBench: Evaluating AI Agents as Market Participants

Dissecting AI Trading: Behavioral Finance and Market Bubbles

Information Aggregation with AI Agents