Markets of Minds: A 51-Billion-State Economy of Intelligence

Jusheng Zhang; Xuekun Wu; Wenhao Wang; Mark Wang; Ming-Hsuan Yang

cs.MAOct 23, 2026

Markets of Minds: A 51-Billion-State Economy of Intelligence

Authors: Jusheng Zhang, Xuekun Wu, Wenhao Wang, Mark Wang, Ming-Hsuan Yang

Organizations: Stanford University · Vast Intelligence Lab · University of California at Merced

Abstract

When intelligence is shared, reasoning becomes a constrained allocation problem. In a population of agents with limited access to expert models, an intervention can benefit its recipient yet still be a poor use of shared capacity if another feasible intervention has greater expected value. Assistance must therefore be evaluated through both its marginal benefit and the opportunity cost induced by budgets, access, scarcity, and service capacity. To this end, we introduce Markets of Minds, a framework for allocating expert reasoning across agents before expert outputs are observed. For each feasible agent-service pair, the framework estimates the expected gain from assistance relative to the agent's default continuation, explicitly separating repair benefit from regression risk, and jointly clears assistance contracts under shared resource constraints. We then study this problem in a dependency-aware agent economy with 51.2 billion addressable persona-market states spanning capabilities, cognitive budgets, provider access, scarcity, and network roles. These dependencies couple allocation feasibility across agents rather than the outcome of a fixed contract. Empirically, in a matched replay of 240 cases under a common 30% assistance quota, the frozen allocator reaches 85.42% task-level persona adherence versus 73.86% for equal-capacity random allocation, a gain of 11.56 percentage points. The results show that assistance value is heterogeneous and can be negative, and that scarcity changes which interventions are worth purchasing. Together, these findings motivate a population-level view of intelligence where the value of assistance depends not only on what it can achieve, but also on which alternative uses of scarce cognitive capacity it displaces.

Figures & tables

Figure 1: Allocation and verification. A stylized execution and analytical illustrations of procurement, verification supply, and net correction.

Figure 2: Markets of Minds. (a) Task-agent execution. (b) Pre-execution value estimation, joint allocation, and resource accounting.

Method	Chat	Web	OS-App	Overall $\uparrow$	Expert use	Cost/succ. $\downarrow$	Harmful help $\downarrow$
No assistance	68.75	61.25	65.00	65.00	0%	$0.06211	–
Random	77.40	70.59	73.60	73.86	30%	$0.06090	0.42%
Difficulty/uncertainty	83.75	76.25	85.00	81.67	30%	$0.05636	0.00%
AutoMix	82.50	76.25	81.25	80.00	30%	$0.07765	0.00%
RouteLLM	86.25	77.50	80.00	81.25	30%	$0.06219	0.00%
GraphRouter	86.25	77.50	80.00	81.25	30%	$0.06023	0.00%

Table 1: Matched allocation at 30% shared expert capacity. 80 cases and 24 assistance slots per environment; cost/success includes method-specific overheads.

Figure 3: Main routing results and budget robustness. Task-level gains, score versus expert use, performance across four expert-use budgets, and the corresponding four-budget mean.

Single-model references
Method	MMMU	MMBench	Math Vision	InfoVQA	CC-OCR	MMMU-Pro	Avg.	Expert use
Gemini 3.5 Flash	76.11	92.99	86.51	78.90	71.42	78.03	80.66	—
Gemini-2.5-Pro	81.70	88.30	63.50	81.00	73.00	68.00	75.92	—
Gemini-2.0-Flash	70.70	83.00	41.30	83.20	73.10	57.00	68.05	—
GPT-5.6 Sol	82.22	83.90	84.54	73.58	79.62	79.77	80.60	—
GPT-4o (2024-08-06)	70.70	74.30	30.40	68.70	66.60	54.50	60.87	—

Table 2: Multimodal benchmark results. Routing entries report three-run means; bold marks the best routing result in each column.

Figure 4: Population-level allocation under shared capacity. The bar charts show welfare under 10%, 30%, and 50% global capacity and their average; the bubble charts show welfare relative to harmful help, Jain fairness, and total API cost.

Variant	Success $\uparrow$	$\Delta_{\mathrm{full}}$ (pp)	Cost/succ. $\downarrow$	Survey $\uparrow$
Restricted model
Priors only	74.22	$-8.44$	$0.0265	70.15
Leave-one-component-out
w/o market prior	79.84	$-2.82$	$0.0252	74.50
w/o uncertainty	81.15	$-1.51$	$0.0241	76.10
w/o preference conflict	80.32	$-2.34$	$0.0245	75.25

Table 3: Cognition allocator ablation. Three-seed mean.

\Delta_{\mathrm{full}}

denotes the success gap relative to the full model in percentage points.

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Allocation rule	Success (%) $\uparrow$	Cost / success $\downarrow$	Harmful help $\downarrow$	Worst group (%) $\uparrow$	Jain $\uparrow$	Missed repairs $\downarrow$
Threshold (arrival order)	78.10	$0.0650	1.20%	71.05	0.965	45
FCFS (positive score)	77.50	$0.0662	1.45%	70.80	0.960	52
Pooled top-72	84.90	$0.0560	0.10%	77.20	0.980	12
Ours (top-24 / environment)	85.42	$0.05509	0.00%	78.75	0.988	15

Appendix

Table 4: Allocation-rule diagnostic. All rules use a nominal 72-slot assistance supply; pooled top-72 relaxes the per-environment capacity constraint.

Predicted-gain quartile	Mean $\widehat{\Delta}$
Top 25% ( $n=100$ )	$0.65$
Second 25% ( $n=100$ )	$0.35$
Third 25% ( $n=100$ )	$0.12$
Bottom 25% ( $n=100$ )	$-0.15$

Appendix

Table 5: Prospective 400-request diagnostic. The left panel summarizes pre-execution value estimates over the full request stream; the right panel reports realized outcomes for the selected subset.

Panel of Figure 5 — Figure 5: Representative qualitative allocation cases. The top row illustrates positive-value assistance and service selection; the bottom row illustrates regression avoidance and a shared failure. Panels (c) and (d) are readability crops of the original benchmark images; the underlying exported images are unchanged. The cases were selected post hoc for diagnosis and do not estimate benchmark prevalence.

Case	Ours	Agora (matched)	RouteLLM (matched)
2 / MMBench	Gemini 3.1 Pro (expert) $\rightarrow$ D $\checkmark$	Gemini 3.5 Flash (base) $\rightarrow$ C $\times$	Gemini 3.5 Flash (base) $\rightarrow$ C $\times$
4 / MathVision	GPT-5.6 Sol (expert) $\rightarrow$ D $\checkmark$	Gemini 3.5 Flash (base) $\rightarrow$ score 0 $\times$	Gemini 3.1 Pro (expert) $\rightarrow$ C $\times$
9 / InfoVQA	Gemini 3.5 Flash (base) $\rightarrow$ 65% $\checkmark$	Claude Sonnet 5 (expert) $\rightarrow$ 54% $\times$	Qwen2.5-VL-72B (expert) $\rightarrow$ 54% $\times$
10 / InfoVQA	Gemini 3.5 Flash (base) $\rightarrow$ rough ER $\times$	Claude Sonnet 5 (expert) $\rightarrow$ plasma membrane $\times$	Qwen2.5-VL-72B (expert) $\rightarrow$ nucleus $\times$

Appendix

Table 6: Allocation traces for the four cases in Figure 5 . “Base” means no expert service was purchased. A check mark denotes a benchmark score of 1.0; a cross denotes 0.0.

Provider access regime	Low income	Lower- middle	Middle	Upper- middle	High income	Premium/ unbounded (%)
	Socio-economic budget level (mean score)
Avoidant	1.10	2.20	3.50	4.10	5.20	11.08%
Cautious adopter	1.35	2.45	3.85	4.80	6.05	13.76%
Comfortable	1.55	2.70	4.25	5.90	7.15	18.92%
Digital native	1.77	2.67	4.24	5.83	7.21	18.87%
Premium/unbounded access (%)	0.53%	0.53%	2.19%	14.40%	46.84%	–

Appendix

Table 7: Dependency sampling diagnostics. The allocation explicitly models correlations between socio-economic status and technological access rather than uniform random assignment.

Method	50%	30%	20%	10%	Avg. welfare $\uparrow$	Harmful help $\downarrow$	Cost / success $\downarrow$
No assistance	73.21	73.21	73.21	73.21	73.21	0.00%	$0.0212
Random	73.28	73.08	73.10	73.15	73.15	14.68%	$0.0256
Difficulty / uncertainty	76.06	76.08	76.53	75.16	75.96	15.45%	$0.0242
Benefit-per-cost greedy	77.40	76.55	75.80	74.50	76.06	11.20%	$0.0225
Agora quality-first	75.48	74.00	74.10	73.98	74.39	8.42%	$0.0240
Ours (joint allocation)	77.90	76.98	76.57	76.08	76.88	2.15%	$0.0218

Appendix

Table 8: 512-agent allocation under shared capacity. Results are reported for the expanded 512-agent diagnostic under 10%, 20%, 30%, and 50% global capacity limits.

Agents	Slots@10%	Slots@30%	Welfare@10%	Welfare@30%	Jain Fair.	Broker Time	Allocations/s	Peak RAM
32	3	10	77.23 $\pm$ 0.69	81.40 $\pm$ 1.02	0.659	6.70 ms	1,503	0.07 MiB
128	13	38	76.10 $\pm$ 2.34	76.77 $\pm$ 1.42	0.855	6.79 ms	4,887	0.25 MiB
512	51	154	78.27 $\pm$ 0.91	80.13 $\pm$ 1.58	0.835	18.12 ms	7,381	1.02 MiB
2,048	205	614	78.10 $\pm$ 0.19	79.10 $\pm$ 0.22	0.863	137.33 ms	4,549	4.23 MiB
10,000	1,000	3,000	78.15 $\pm$ 0.12	79.25 $\pm$ 0.15	0.871	785.40 ms	12,732	20.80 MiB
100,000	10,000	30,000	78.20 $\pm$ 0.05	79.31 $\pm$ 0.08	0.875	8,450.15 ms	11,834	210.55 MiB

Appendix

Table 9: System scalability and computational overhead across increasing population sizes. The allocation mechanism scales efficiently to 100,000 concurrent agents within an 8.5-second clearing window.

Multi-Agent Resource Allocation

See all 28 papers

Jun 1, 2026

Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions

How can a population of agents self-orchestrate and self-adapt into stronger collective intelligence without centralized control? Inspired by Friedrich Hayek's economic theory of decentralized coordination in markets, we study this question through an agent economy in which agents compete via auctions for the right to act, exchange payments, and accumulate wealth from environmental rewards. These simple local incentives induce decentralized credit assignment, driving global planning without explicit orchestration or communication protocols. The population evolves through economic selection: effective agents accumulate wealth and are mutated via exploitation, while ineffective ones go bankrupt and are replaced via exploration. Using modern Large Language Model (LLM) based agents, we show that, initialized with weak agents, the agent economy produces emergent multi-step reasoning strategies and outperforms stronger monolithic baselines across five agentic tasks, including mathematical reasoning, financial research, scientific research, accelerator design, and distributed-system optimization. Our results suggest a new path to multi-agent intelligence: rather than engineering coordination, we can design decentralized incentive structures under which it automatically emerges.

Zhenting Qi, Huangyuan Su, Ao Qu +13

Harvard · MIT

Multi-Agent Coordination Multi-Agent Reasoning Self-Evolving Agents Emergent Behavior in Multi-Agent Systems

Ask PDF

Aug 24, 2026

Markets, Not Planners: Decentralized Orchestration of LLM Agents with Private Information

As LLM agents proliferate, built by different parties and with different capabilities and costs, orchestrating them is more like assembling labor across the economy than a computer calling a subroutine. Existing orchestration is typically centralized, with a single planner assigning every task, but this creates a bottleneck as agent pools grow, requires private information (e.g., agents' execution costs), and can easily be manipulated, such that a single inserted preference nearly doubles a favored agent's task share under a centralized LLM allocator. We introduce AgentLance, a repeated labor market in which agents bid on tasks using their private costs and self-maintained strategy notes, an allocator selects winners from bids and public reputation records, and a VCG-style payment rule rewards cost-aware bidding. Complex tasks are handled by hierarchical delegation: winning agents can decompose work and subcontract it through the same mechanism. Across mathematical reasoning, code generation, knowledge-intensive QA, and agentic tasks, AgentLance matches agents to their specializations, shifts work toward cheaper agents as cost sensitivity rises, and consistently outperforms single-model, centralized-orchestration, and market baselines. Diagnosing market failures, including inaccurate cost self-estimation and sub-optimal bidding, then correcting them in controlled experiments yields further gains, charting a path toward more efficient agent economies.

Xiao Liu, Haoyang Li, Songwei Li +4

Knowledge Lab, University of Chicago · University of Virginia · Tsinghua University +2

Multi-Agent Orchestration LLM Agent Orchestration Multi-Agent Resource Allocation Mechanism Design

Ask PDF

Aug 17, 2026

$R^3$ -Bench: LLMs Struggle with Resource-Rational Reasoning under Shared Budgets

In cognitive science, resource rationality asks how an agent should allocate limited computation to maximize expected value. Most reasoning and agent benchmarks use independent per-task budgets; existing shared-budget studies do not calibrate suite performance against the same model's demonstrated single-problem competence. We introduce

R^3

-Bench, which evaluates six-problem suites under shared budgets across mathematics, competitive programming, and abstract reasoning in tool-free and agentic settings. Matched single-problem response curves define an offline empirical oracle over observed successes. Across 72 main-table cells for six models, the oracle mean matches or exceeds the contest mean in all cells and is strictly higher in 71. Under moderate tool-free pressure, equal-allocation replay also exceeds contest performance for four of six models. Trajectory diagnostics reveal limited strategy updating and pressure-dependent failure patterns. In a three-model diagnostic under strong agentic pressure, at least one fixed scheduler exceeds the contest mean in six of nine cells, but no policy dominates across domains. These results expose a persistent gap between demonstrated competence and shared-budget realization.

Peisong Wang, Zhiwei Ma, Bowen Liu +6

The Hong Kong University of Science and Technology (Guangzhou) · The University of Hong Kong · Hunyuan Team, Tencent

LLM Reasoning AI Agent Benchmarks Bounded Rationality Benchmark Design

Ask PDF

Abstract

Multi-Agent Resource Allocation

Explore similar work

Economy of Minds: Emerging Multi-Agent Intelligence with Economic Interactions

Markets, Not Planners: Decentralized Orchestration of LLM Agents with Private Information

R3R^3R3-Bench: LLMs Struggle with Resource-Rational Reasoning under Shared Budgets

$R^3$ -Bench: LLMs Struggle with Resource-Rational Reasoning under Shared Budgets