Eco3S: Complex Socio-Economic System Simulation via Agent-Based Models

Gang Kou; Shaopeng Wei; Wenxi Sun; Yepeng Ding; Yufei Cheng; Yu Zhao

REVIEW 4 major objections 5 minor 23 references

The paper argues that an agent-based simulator with a co-evolving environment can make LLM-driven agents reproduce established economic findings, from canal-decay rebellion to information spread.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · deepseek-v4-flash

2026-08-01 12:44 UTC pith:QZAUMRVE

load-bearing objection A well-built integration of LLM-ABM components whose headline replications are undercut by calibration-to-target and probable LLM memorization; worth sending to reviewers but with expectations of major revision. the 4 major comments →

arxiv 2607.26588 v1 pith:QZAUMRVE submitted 2026-07-29 cs.AI

Eco3S: Complex Socio-Economic System Simulation via Agent-Based Models

Shaopeng Wei , Yufei Cheng , Wenxi Sun , Yepeng Ding , Yu Zhao , Gang Kou This is my paper

classification cs.AI

keywords Eco3Sagent-based modelingLLM agentsco-evolving environmentstructural causal simulationcounterfactual reasoningautomatic simulation refinementeconomic history replication

verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper tries to establish that agent-based economic simulation becomes scientifically useful when the environment is not a static backdrop: Eco3S lets physical, social, and economic subsystems change in response to agent decisions, so macro outcomes like rebellion, governance, and information spread emerge from micro behavior. Three mechanisms do the work: a bidirectional co-evolving environment, an SCM-style counterfactual mechanism that replays simulations from saved checkpoints under different interventions, and an automated refinement loop that turns natural-language research prompts into executable simulations. The authors test this on three published economic studies (canal decay and rebellion, origins of governance, information delivery). They report close matches: simulated rebellion risk differs from the published effect by 8 percentage points, and the spatial canal/non-canal rebellion gap (125%) aligns with the empirical 117%. A sympathetic reading: if right, Eco3S offers a way to stress-test economic and policy hypotheses before implementation in a controllable, counterfactual-rich setting.

Core claim

On the paper's own terms, Eco3S demonstrates that LLM-driven agents, when coupled to a co-evolving environment through the transition equations (1), produce emergent macro-dynamics that match established findings. In the canal-decay scenario, declining navigability raises unemployment, lowers satisfaction, and concentrates rebellion in canal-side towns; the reported 0.54 vs 0.24 rebellion rates give a 125% increase against the 117% empirical benchmark, and the overall rebellion-risk deviation is 8 percentage points. Structural Causal Simulation then quantifies the effects of removing sea transport, maintenance, or climate shocks by replaying the same baseline state from checkpoints, finding

What carries the argument

Co-evolving Environment Design, the paper's named central mechanism, is a bidirectional feedback loop: agents perceive environmental states, decide, and their aggregated actions reshape the physical, social, and economic environment, which in turn changes the next round of decisions (formalized in equation (1)). Structural Causal Simulation adds an SCM-inspired do-operator: the simulator keeps checkpoints and can resume from the same state under different policies, producing paired trajectories whose difference is treated as a causal effect. A third mechanism, the Simulation-Analysis-Refinement (SAR) paradigm, uses four LLM agents to turn a natural-language request into runnable Python simul

Load-bearing premise

The load-bearing premise is that agents decide based on the simulated state the framework feeds them, not on historical knowledge the language model already carries from its training data; if those prompts leak the outcome, the replication validates the model's memory rather than the simulator.

What would settle it

Take the exact canal-decay configuration but rename all places and eras, strip historical proper nouns and outcome hints from the agent prompts, and keep only the environmental variables. If the rebellion pattern does not concentrate along the decaying waterway, the co-evolving environment is not producing the result; if it does, the framework passes.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

If the canal-decay replication holds, the causal chain infrastructure decay → unemployment → falling satisfaction → rebellion is validated as an emergent outcome, and policy interventions (maintenance, climate, sea transport) can be compared by re-running from the same checkpoint.
If the origins-of-governance case holds, shared infrastructural need is sufficient to make state affiliation an economically rational individual choice, supporting the cooperative theory of state formation.
If the information-delivery replication holds, common knowledge has opposite effects under broadcasting versus seeding, which can guide how governments and organizations publicize information.
If the auto-simulation results hold, researchers can obtain working simulations of herding, hysteresis, asset bubbles, and segregation from high-level prompts within roughly five refinement cycles, with scale-up to 10,000 agents at manageable cost.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

A test the paper does not run would be decisive: swapping the historical setting for a fictional geography with the same formal canal-decay structure. If rebellion no longer tracks infrastructure decline, the benchmark match reflects the LLM's training-data prior, not Eco3S's co-evolution.
The same checkpoint/replay machinery could be used as a low-cost ex-ante policy lab for settings with no historical benchmark, but only if simulation outputs are validated against external causal estimates; otherwise the counterfactual contrasts risk being differences in LLM role-play rather than policy effects.
Because the framework's validity ultimately rests on LLM reasoning quality, smaller or weaker models (the paper itself notes one model diverges) should be treated as a boundary condition; a practical extension is to run all reported experiments with a 'blind' LLM that is not told it is simulating history.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Desk Editor's Note

A well-built integration of LLM-ABM components whose headline replications are undercut by calibration-to-target and probable LLM memorization; worth sending to reviewers but with expectations of major revision.

read the letter

Alex,

The thing to know: Eco3S is a serious attempt to build one platform out of three ideas—co-evolving environment-agent feedback, SCM-inspired checkpoint counterfactuals, and an LLM-driven auto-simulation loop. The integration is real and the engineering is careful. The paper deserves credit for the detailed appendix, the ablations, the cross-provider robustness test, and the willingness to state its own limitations. Given how much of this literature is demo-ware, this is better than typical.

But the validation does not yet support the replication claims. Two issues are load-bearing.

First, the results are partly fitted to the target. Appendix A.1 says final settings were chosen for 'stability of the principal qualitative trends,' and the SAR loop terminates when outputs show 'plausible trends' (K≤10). That is a search over parameters and designs with the benchmark outcome known in advance. It can produce direction-matching, but it isn't an out-of-sample test.

Second, the LLM-memorization worry is real. The Canal Decay prompts position agents as Qing-era residents reacting to canal decay, taxation, and rebellion—content that is in any LLM's training data, and the benchmark paper (Cao & Chen 2022) is a known historical economics result. The Appendix D robustness check only swaps providers, all of which share Chinese-history pretraining. It cannot distinguish 'the framework generated the pattern' from 'the LLM retrieved the familiar narrative.' The paper's own Discussion concedes exactly this and proposes cross-configuration convergence testing; the current experiments do not include it.

There are smaller issues. Table 3's Choice Quality row shows a simulated +1.3% versus the field's +81.0%—calling that a replication is generous, even with the 'similar directional patterns' framing. The code and full prompts are not in the version I read; the appendix says they will be released. And Eq. (3) is a simple percentage-change comparison, not a serious SCM contribution—but that is a minor terminological inflation, not a flaw in the platform.

Net: the framework is a plausible starting point for LLM-based economic simulation, and the authors are honest about its gaps. But the headline quantitative alignments should not be taken at face value. If the artifact is released and an out-of-sample or decontextualized ablation lands, this could become a much stronger paper. I'd send it to peer review—it deserves a serious referee—and ask for those revisions. I'd cite it as a platform example, not for the numbers.

Referee Report

4 major / 5 minor

Summary. The paper proposes Eco3S, an LLM-driven agent-based modeling framework for socio-economic simulation, with three claimed innovations: a co-evolving agent-environment feedback loop, an SCM-inspired counterfactual mechanism, and a Simulation-Analysis-Refinement (SAR) pipeline that auto-generates and refines simulations from natural-language prompts. Validation is attempted through three benchmark replications—Canal Decay and Rebellion (Cao and Chen 2022), Origins of Governance (Allen et al. 2023), and Information Propagation (Banerjee et al. 2024)—plus additional auto-simulation case studies (herding, hysteresis, asset bubbles, Schelling segregation), ablations, scalability experiments, and a comparison with existing platforms. The central claim is that Eco3S reproduces established empirical findings and that the observed emergent patterns are driven by the framework's coupled dynamics rather than by the LLM's prior knowledge.

Significance. If the central claim were established, Eco3S would be a useful contribution to LLM-based ABM: it addresses a real gap (evolving physical/social environments), provides a structured causal-intervention layer, and offers an automation pipeline that lowers the barrier to simulation-based economic research. The paper is commendably transparent in some respects: Appendix A.1 documents parameter-search criteria, Appendix A.3 states which comparisons are descriptive and which are inferential, and the Discussion explicitly acknowledges that validation without historical benchmarks remains an open problem. The robustness check across several LLM providers (Appendix D) is a useful sanity check. However, the current evidence does not yet separate the framework's generative contribution from the LLM's memorized historical narrative, and the reported quantitative matches are partly based on metric comparisons that overstate agreement. The proposed framework and its limitations are clearly presented, and the key missing experiments are concrete and feasible.

major comments (4)

[Canal Decay and Rebellion; Appendix A.4; Appendix D; Discussion] The most load-bearing validity threat is LLM memorization. The agents are prompted with historically specific roles and context—Qing-era canal decay, transport costs, taxation, and a five-level satisfaction scale ending in "hate the regime, vow to overthrow it"—and the benchmark (Cao and Chen 2022) is a well-known historical-economics result that is very likely present in the pretraining data. The robustness check in Appendix D only swaps LLM providers; it does not test whether the rebellion pattern persists when the same environmental rules are described in historically neutral or anonymized terms. The paper's own Discussion admits that cross-configuration convergence testing is needed and is not performed. A decontextualized-prompt ablation, or an ablation that removes the canal-decay-to-cost-to-unemployment pathway, is required to support the claim that the simulated recurrence emerge
[Table 3 and text 'Information Propagation'] The text states that the simulated results show "similar directional patterns" and "closely aligns" with Banerjee et al. (2024), but Table 3 shows a very uneven quantitative match. For Choice Quality, the field effect is +81.0% while Eco3S gives +1.3%, i.e., essentially no effect; for Knowledge, the simulation overshoots by a factor of roughly 7 (+38.5% vs +5.6%); for Conversation under Seeding, the simulation overshoots by a factor of 7.5 (+773.7% vs +103.0%). Only the Broadcast/CK conversation reduction (+64.6% vs +63.0%) is quantitatively close. Calling this a replication of the field experiment requires a pre-specified agreement metric and error bars; as reported, the evidence supports at most a partial directional match, not the paper's "replicates established economic studies" claim.
[Appendix A.1; Auto-simulation; Simulation-Analysis-Refinement] The validation pipeline has two degrees of freedom that can inject the target pattern. Appendix A.1 states that final settings were selected for "stability of the principal qualitative trends," and the SAR loop terminates when outputs "exhibit plausible trends" or reach K≤10, with the ResearchAnalystAgent diagnosing and adjusting configurations after each run. Since the target empirical outcomes are known before configuration, the reported results are not out-of-sample predictions. The paper needs to report, for each replication, the number of refinement iterations, the specific parameter/prompt changes made, and a sensitivity analysis in which the SAR loop and manual tuning are disabled (e.g., using default or pre-registered parameters). Without this, the headline "8 percentage point deviation" cannot be distinguished from fitting to the target.
[Canal Decay paragraph; Origins of Governance] The quantitative claim in the Canal Decay experiment is not precisely defined. The text says "the simulated rebellion risk deviates from the reported effect by only 8 percentage points," but what is actually compared later is a simulated +125% spatial gap against the empirical +117% gap—both percentage increases. A difference between two relative-increase measures is not a "rebellion risk" deviation. Similarly, the Origins of Governance experiment reports no quantitative benchmark at all; it concludes alignment with Allen et al. (2023) based on qualitative trends in river navigability and urban population. If the paper's contribution is presented as replicating established economic studies, each replication needs a pre-specified target metric, the simulated value, and an explicit comparison with uncertainty, rather than a narrative alignment.

minor comments (5)

[Figure 3 caption] Typo: "Resisdent Satisfaction" should be "Resident Satisfaction."
[Table 1 and text] The text says "p < 0.05" for the canal vs. non-canal comparison, while Table 1 reports p < 0.01. Please reconcile.
[Table 3 column header] The column header "S BC Cons." is unclear. The table reports Seeding and Broadcasting conditions, but the arrangement of conditions and the meaning of the checkmark column should be explained in the caption or footnotes.
[Appendix C] The figures in Appendix C (Figures 10–13) appear with garbled axis labels and textual artifacts, making the auto-simulation results difficult to verify. The figures should be regenerated or the source data should be provided.
[Appendix A.3 and code availability] The paper mentions that analysis code documents the processing of run-level outputs, but does not provide a working repository link. Since reproducibility is a stated strength, a stable and complete artifact link should be included.

Circularity Check

3 steps flagged

Canal Decay 'replication' is calibrated reproduction: trend-based parameter selection, SAR satisficing, and hard-coded canal→unemployment→satisfaction→rebellion rules load the target pattern into the simulation.

specific steps

fitted input called prediction [Auto-simulation — Results Analysis and Configuration Optimization]
"Rather than seeking a strict mathematical optimum, SAR employs a diagnostic-guided satisficing principle; the process terminates once the results exhibit plausible trends or reach the iteration limit, typically 10. Empirically, optimization usually succeeds within five cycles. The resulting calibrated simulation serves as a baseline for subsequent counterfactual interventions."

The SAR loop iteratively reconfigures code, parameters, and prompts until the ResearchAnalystAgent judges results satisfactory; the termination condition is 'plausible trends.' Therefore auto-simulation 'replications' (herding, hysteresis, bubbles, Schelling) are outputs fitted to the expected pattern, not independent predictions. Using the calibrated simulation as baseline then carries that target pattern into all counterfactual comparisons.
self definitional [Appendix A.4 'Rebel on the Canal' — Environment and Agents]
"The canal transport cost is defined as the baseline cost multiplied by (2−navigability index) ... Maritime transport is fixed at one-fifth of the baseline cost ... Canal maintenance also directly affects the number of available infrastructure maintenance positions in the labor market. ... Resident satisfaction is represented on a five-level qualitative scale: 'hate the regime, vow to overthrow it'; ... This categorical measure directly influences the probability of joining a rebellion."

The canal-decay-to-rebellion relationship is an input rule, not an emergent output. Navigability loss is defined to raise transport cost and eliminate maintenance jobs; residents are defined with satisfaction levels that include 'hate the regime' and this level 'directly influences the probability of joining a rebellion.' The observed sequence of declining navigability, unemployment, dissatisfaction, and rebellion is the implemented mechanism, so the benchmark agreement with Cao and Chen (2022) is a check of the hard-coded rules.
fitted input called prediction [Appendix A.1 Parameter Development; Experiments — Canal Decay and Rebellion]
"We conducted non-exhaustive pilot tests rather than optimizing parameters against the reported evaluation metrics. Final settings were selected by jointly considering the stability of the principal qualitative trends ... Benchmarked against Cao and Chen (2022), the simulated rebellion risk deviates from the reported effect by only 8 percentage points."

Final settings were selected using the target qualitative trend as a criterion. The headline 8 percentage-point deviation and the 125%-vs-117% spatial-gap comparison are computed with these tuned settings, making them in-sample goodness-of-fit statistics rather than out-of-sample predictions. Without a holdout configuration or a decontextualized-prompt ablation, the numbers cannot distinguish framework-driven emergence from calibrated reproduction.

full rationale

The paper's central replication claims reduce to their own inputs. (1) A.1 chooses final settings for 'stability of the principal qualitative trends,' so the quantitative closeness of the Canal Decay results is fitted rather than predicted. (2) The SAR auto-simulation loop terminates only when outputs 'exhibit plausible trends' and is explicitly described as producing a 'calibrated simulation.' (3) A.4 hard-codes the full causal pathway from canal navigability loss to transport cost, maintenance-job loss, dissatisfaction, and rebellion, including a satisfaction scale whose lowest category 'directly influences the probability of joining a rebellion.' Thus the simulated 8pp deviation and 125% spatial gap are consequences of the implemented rules and tuning criteria, not emergent discoveries. The LLM-memorization confound is a further validity threat: prompts describe Qing-era canal politics and rebellion, all providers in Appendix D share pretraining on this history, and the paper does not run a decontextualized-prompt ablation. The Discussion itself concedes the problem: 'Evaluating simulation validity without historical benchmarks remains challenging,' proposing cross-configuration convergence testing that is not carried out. I found no load-bearing self-citation chain; external benchmarks (Cao and Chen 2022, Allen et al. 2023, Banerjee et al. 2024) are real. But the derivation of the headline replication is circular enough to warrant score 6: partial circularity by construction.

Axiom & Free-Parameter Ledger

8 free parameters · 5 axioms · 0 invented entities

The central simulations rest on hand-selected parameters and on the assumption that LLM behavior faithfully represents human agents. No independent evidence is provided for these assumptions beyond qualitative matching of known trends.

free parameters (8)

Response probability = 0.05 (historical); 0.2 (Information Delivery)
Controls the fraction of agents that act each step; selected in Table 5 for LLM-call cost and qualitative trend stability, affecting all emergent dynamics.
LLM temperature = 1.0
Selected as the highest tested value to encourage agent decision diversity while retaining stable aggregate behavior; directly shapes behavior variability.
Population size = 2,000 (historical); 200 (Information Delivery)
Chosen to balance aggregate stability and API cost; qualitative outcomes may depend on this scale.
Simulation horizon = 15 (Canal); 10 (Origins); 3 (Information Delivery)
Set to cover the temporal dynamics of each benchmark, not derived from the benchmark data.
Canal navigability decay and climate factors = δ, γ in φ_{t+1}=max(0, φ_t·(1−δ)−γ·0.6)
Hand-set environmental dynamics; no data fitting is reported, and these rules push the canal toward decline in every scenario.
Transport cost multipliers = canal cost = baseline × (2−navigability); maritime = 0.2 × baseline
Hand-set to create the economic incentive to shift from canal to sea, encoding the very trade-disruption mechanism the Canal Decay experiment claims to reproduce.
Conflict attrition parameters = baseline c=0.1, decay d∈[0.5,1.0], noise ε ~ U(−0.03,0.03)
Hand-set in A.4 to define rebellion losses; no independent estimation is given.
HIN power-law exponent = a in f(x)=c x^{-a}
Social network degree distribution is generated with a power law whose exponent is chosen by hand, affecting all information diffusion and rebellion propagation.

axioms (5)

domain assumption LLM outputs faithfully approximate human decision-making in historical and economic contexts.
All agent behavior is mediated by LLM prompts; no calibration or validation of this psychological assumption is provided.
domain assumption Pretrained LLM knowledge does not leak the expected historical outcome into agent decisions.
The Canal and Origins simulations use historically named settings whose outcomes are known from training data; the paper does not control for knowledge contamination.
domain assumption A few hundred to a few thousand agents adequately represent a population for the reported macro conclusions.
Results use 200 or 2,000 agents; no demographic or sampling justification is given beyond variance reduction.
domain assumption The benchmark studies' reported effects are faithfully captured by the paper's simplified comparison metrics.
For example, a 125% simulated contrast is compared to a 117% empirical benchmark without showing that the underlying quantities measure the same estimand.
domain assumption Equation (1) adequately formalizes agent-environment co-evolution, and checkpoint replay in Equation (3) yields interpretable causal effects.
The transition functions F_a and F_e are not instantiated rigorously in the paper, and Equation (3) is a simple difference of simulation outputs with no identification conditions.

pith-pipeline@v1.3.0-daily-deepseek · 36864 in / 13210 out tokens · 128625 ms · 2026-08-01T12:44:45.570307+00:00 · methodology

0 comments

read the original abstract

The rapid development of large language models (LLMs) has renewed interest in agent-based modeling (ABM). However, current LLM-based ABM research faces several key challenges: modeling evolving agent-environment interactions, enabling flexible counterfactual reasoning, and automating simulation workflows for scientific research. In this paper, we propose Eco3S, a socio-economic system simulation framework for economic research and policy analysis that addresses these challenges through three key mechanisms: (1) Co-evolving Environment Design, a bidirectional feedback loop where agents and the environment co-evolve, producing realistic emergent behaviors; (2) Structural Causal Simulation, a structural causal model (SCM)-inspired counterfactual mechanism that allows flexible interventions for diverse causal inference tasks; (3) Simulation-Analysis-Refinement Paradigm, a self-corrective mechanism that iteratively refines experimental designs based on prior simulation results. Experiments on diverse economic scenarios confirm \textit{Eco3S}'s effectiveness in replicating multiple established economic studies (canal decay, origins of governance, and information propagation) and phenomena across domains. Additional results further demonstrate its scalability and generalizability, highlighting the framework's potential for rigorous economic research and policy-making.

Figures

Figures reproduced from arXiv: 2607.26588 by Gang Kou, Shaopeng Wei, Wenxi Sun, Yepeng Ding, Yufei Cheng, Yu Zhao.

**Figure 1.** Figure 1: The architecture of the proposed Eco3S framework. Basic Framework Agent Configuration. The configuration module defines the setup for agents, the environment, and their interaction rules. Each agent profile includes static attributes and dynamic states, paired with a dual short- and long-term memory system. To handle LLM context limits, a memory summarization mechanism bounds the context length while pre… view at source ↗

**Figure 2.** Figure 2: The architecture of the proposed auto-simulation [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 5.** Figure 5: Stacked Distribution of Rebellions. Origins of Governance Drawing on Allen, Bertazzini, and Heldring (2023), we investigate the cooperative theory of state formation. Specifically, shifting river patterns can render private irrigation unsustainable, thereby driving the demand for an authority to coordinate public canal infrastructure. Eco3S implements this by modeling a feedback loop between the river s… view at source ↗

**Figure 4.** Figure 4: Comparison of Spatial Distribution of Rebellion. [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗

**Figure 6.** Figure 6: Results of the Origins of Governance experiment. [PITH_FULL_IMAGE:figures/full_fig_p005_6.png] view at source ↗

**Figure 7.** Figure 7: Results of the Information Delivery experiment. [PITH_FULL_IMAGE:figures/full_fig_p006_7.png] view at source ↗

**Figure 8.** Figure 8: Benchmark comparison for Canal Decay. Figures 8(a) and (b) indicate that all platforms capture the decline in navigability, yet their labor market responses diverge. In Eco3S, unemployment rises alongside canal deterioration, while YuLan-OneSim and GenSim exhibit weaker or logically inconsistent responses. This disparity underscores the efficacy of Co-evolving Environment Design, which couples environme… view at source ↗

**Figure 9.** Figure 9: Ablation study of the Canal Decay experiment. [PITH_FULL_IMAGE:figures/full_fig_p007_9.png] view at source ↗

**Figure 12.** Figure 12: Results of Asset Bubble Experiment. C.4 Schelling Segregation Model This study reconstructs Schelling (1971)’s model on a grid to observe macro-level spatial segregation. Agents relocate if neighbor similarity falls below a threshold. Segregation is measured by the Index of Dissimilarity (D): D = 1 2 Xn i=1 [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗

**Figure 11.** Figure 11: Results of Custom Satisfaction Experiment. [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗

**Figure 13.** Figure 13: Results of Schelling Model Experiment [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗

**Figure 14.** Figure 14: Consistency of simulation trends across different [PITH_FULL_IMAGE:figures/full_fig_p015_14.png] view at source ↗

**Figure 17.** Figure 17: Runtime interface for the multi-phase simulation [PITH_FULL_IMAGE:figures/full_fig_p016_17.png] view at source ↗

**Figure 18.** Figure 18: Dashboard for managing generated simulation [PITH_FULL_IMAGE:figures/full_fig_p016_18.png] view at source ↗

**Figure 19.** Figure 19: Simulation execution and real-time monitoring [PITH_FULL_IMAGE:figures/full_fig_p017_19.png] view at source ↗

discussion (0)

Reference graph

Works this paper leans on

23 extracted references · 13 linked inside Pith

[2]

Theeconomyneedsagent- based modelling.Nature,

Farmer,J.D.;andFoley,D.2009. Theeconomyneedsagent- based modelling.Nature,

2009
[4]

Unveilingthetruth and facilitating change: Towards agent-based large-scale so- cialmovementsimulation

Mou,X.;Wei,Z.;andHuang,X.-J.2024. Unveilingthetruth and facilitating change: Towards agent-based large-scale so- cialmovementsimulation. InFindingsoftheAssociationfor Computational Linguistics. Park, J. S.; O’Brien, J.; Cai, C. J.; Morris, M. R.; Liang, P.; and Bernstein, M. S

2024
[5]

Gao, S.; Wen, Y.; Zhu, M.; Wei, J.; Cheng, Y.; Zhang, Q.; and Shang, S

S3: Social-network simulation sys- tem with large language model-empowered agents.arXiv preprint arXiv:2307.14984. Gao, S.; Wen, Y.; Zhu, M.; Wei, J.; Cheng, Y.; Zhang, Q.; and Shang, S

Pith/arXiv arXiv
[7]

Jin, H.; Yoo, M.; Park, J.; Lee, Y.; Wang, X.; and Kim, J

War and peace (wara- gent): Large language model-based multi-agent simulation of world wars.arXiv preprint arXiv:2311.17227. Jin, H.; Yoo, M.; Park, J.; Lee, Y.; Wang, X.; and Kim, J

Pith/arXiv arXiv
[11]

InProceedings of the Annual ACM Symposium on User Interface Software and Technol- ogy

Generative agents: Interactive simulacra of human behavior. InProceedings of the Annual ACM Symposium on User Interface Software and Technol- ogy. Pearl, J. 2009.Causality: Models, Reasoning and Inference. Cambridge University Press, 2 edition. Piao, J.; Yan, Y.; Zhang, J.; Li, N.; Yan, J.; Lan, X.; Lu, Z.; Zheng,Z.;Wang,J.Y.;Zhou,D.;etal.2025. AgentSocie...

Pith/arXiv arXiv 2009
[12]

Schelling,T.C.1971

Y social: an llm-powered social me- dia digital twin.arXiv preprint arXiv:2408.00818. Schelling,T.C.1971. Dynamicmodelsofsegregation.Jour- nal of mathematical sociology,

Pith/arXiv arXiv 1971
[13]

InProceedings of the Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (System Demonstrations)

Gensim: A general social simulation platform with large language model based agents. InProceedings of the Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (System Demonstrations). Tang,X.;Zou,A.;Zhang,Z.;Li,Z.;Zhao,Y.;Zhang,X.;Co- han,A.;andGerstein,M.2024. MedAgents:Largelangu...

2024
[14]

Touzel,M.P.;Sarangi,S.;Welch,A.;K,G.;Zhao,D.;Yang, Z.;Yu,H.;Gibbs,T.;Kosak-Hine,E.;Musulan,A.;Thibault, C.; Rabbany, R.; Godbout, J.-F.; and Pelrine, K

Simulating social media using large language models to evaluate alternative news feed algorithms.arXiv preprint arXiv:2310.05984, abs/2310.05984. Touzel,M.P.;Sarangi,S.;Welch,A.;K,G.;Zhao,D.;Yang, Z.;Yu,H.;Gibbs,T.;Kosak-Hine,E.;Musulan,A.;Thibault, C.; Rabbany, R.; Godbout, J.-F.; and Pelrine, K

Pith/arXiv arXiv
[15]

Xu, R.; Lu, D.; Tan, X.; Wang, X.; Yuan, S.; Chen, J.; Chu, W.; and Xu, Y

Simulating public administration crisis: A novel generative agent-based simu- lation system to lower technology barriers in social science research.arXiv preprint arXiv:2311.06957. Xu, R.; Lu, D.; Tan, X.; Wang, X.; Yuan, S.; Chen, J.; Chu, W.; and Xu, Y

Pith/arXiv arXiv
[16]

Yang, C.; Zhang, Y.; Wang, J.; He, L.; and Wu, H

Mindecho: Role-playing lan- guage agents for key opinion leaders.arXiv preprint arXiv:2407.05305. Yang, C.; Zhang, Y.; Wang, J.; He, L.; and Wu, H. 2025a. Adeepreinforcementlearningbasedmulti-agentsimulation optimization approach for IGV bidirectional task allocation and charging joint scheduling in automated container termi- nals.Computers & Operations R...

Pith/arXiv arXiv
[17]

Zhang, X.; Lin, J.; Mou, X.; Yang, S.; Liu, X.; Sun, L.; Lyu, H.; Yang, Y.; Qi, W.; Chen, Y.; et al

OASIS: Open agents social interaction simulations on one million agents.arXiv preprint arXiv:2411.11581. Zhang, X.; Lin, J.; Mou, X.; Yang, S.; Liu, X.; Sun, L.; Lyu, H.; Yang, Y.; Qi, W.; Chen, Y.; et al

Pith/arXiv arXiv
[18]

Zhang, X.; Lin, J.; Sun, L.; Qi, W.; Yang, Y.; Chen, Y.; Lyu, H.; Mou, X.; Chen, S.; Luo, J.; et al

Socioverse: A world model for social simulation powered by llm agents and a pool of 10 million real-world users.arXiv preprint arXiv:2504.10157. Zhang, X.; Lin, J.; Sun, L.; Qi, W.; Yang, Y.; Chen, Y.; Lyu, H.; Mou, X.; Chen, S.; Luo, J.; et al

Pith/arXiv arXiv
[19]

Zhao,K.I.;Naim,M.;Kondic,J.;Cortes,M.E.;Ge,J.;Luo, S.; Yang, G

Electionsim: Massive population election simulation powered by large language model driven agents.CoRR. Zhao,K.I.;Naim,M.;Kondic,J.;Cortes,M.E.;Ge,J.;Luo, S.; Yang, G. R.; and Ahn, A. 2024a. LyfeAgents: genera- tive agents for low-cost real-time social interactions.arXiv preprint arXiv:2310.02172. Zhao, Q.; Wang, J.; Zhang, Y.; Jin, Y.; Zhu, K.; Chen, H.;...

Pith/arXiv arXiv
[20]

AlphaAgents: Large Language Model basedMulti-AgentsforEquityPortfolioConstructions.arXiv preprint arXiv:2508.11152, abs/2508.11152. A Main Experiment Details A.1 Parameter Development and Selection We conducted non-exhaustive pilot tests rather than opti- mizing parameters against the reported evaluation metrics. Finalsettingswereselectedbyjointlyconsider...

Pith/arXiv arXiv
[21]

canal maintenance workers,

Responseprobabilitywasselectedjointlywithpopulation size because the expected number of agent responses, and consequently the number of LLM calls, increases with both quantities. For the historical simulations with2,000agents, Parameter Values Examined Final Setting Selection Criterion Response probability0.05,0.2,0.3,0.4,0.5,0.8 0.05for the historical si...

2024
[22]

Schellingparadox

Trading volume was volatile with notable spikes, and the bid-ask spread narrowed over time. These results demonstrate that LLM-driven agents can en- dogenously generate asset bubbles through speculative trad- ing and decision feedback. /uni00000013/uni00000018/uni00000014/uni00000013/uni00000014/uni00000018/uni00000015/uni00000013/uni00000015/uni00000018/...

1971
[23]

Figure 18: Dashboard for managing generated simulation projects

This interface provides a centralized entry point for inspecting project descriptions and initiating subsequent operations. Figure 18: Dashboard for managing generated simulation projects. During simulation execution, the monitoring interface in Figure 19 presents live charts of key metrics together with a dynamic map of agent locations and states. This s...

2023
[113]

G.; and Golub, B.2024

Banerjee, A.; Breza, E.; Chandrasekhar, A. G.; and Golub, B.2024. WhenLessIsMore:ExperimentalEvidenceonIn- formation Delivery During India’s Demonetisation.Review of Economic Studies,

2024
[645]

PrincetonUniversityPress

Hayashi,F.2011.Econometrics. PrincetonUniversityPress. Hu,G.;He,W.;Sun,C.;Zhu,H.;Li,K.;andJiang,L.2023. Hierarchical belief rule-based model for imbalanced multi- classification.Expert Systems with Applications,

2011
[2022]

Rebel on the canal: Disrupted tradeaccessandsocialconflictinChina,1650–1911.Amer- ican Economic Review,

1911
[2023]

Allen, R

GPT-4 Technical Report.arXiv preprint arXiv:2303.08774. Allen, R. C.; Bertazzini, M. C.; and Heldring, L

Pith/arXiv arXiv
[2024]

Li,C.;Su,X.;Han,H.;Xue,C.;Zheng,C.;andFan,C.2023

Episodic Future Thinking MechanismforMulti-agentReinforcementLearning.InPro- ceedings of the Neural Information Processing Systems. Li,C.;Su,X.;Han,H.;Xue,C.;Zheng,C.;andFan,C.2023. Quantifying the impact of large language models on collec- tive opinion dynamics.arXiv preprint arXiv:2308.03313. Li,J.;Lai,Y.;Li,W.;Ren,J.;Zhang,M.;Kang,X.;Wang,S.; Li,P.;Zha...

Pith/arXiv arXiv 2023
[2025]

InPro- ceedings of the CHI Conference on Human Factors in Com- puting Systems

TeachTune: Reviewing Pedagogical Agents Against Diverse Student Profiles with Simulated Students. InPro- ceedings of the CHI Conference on Human Factors in Com- puting Systems. Association for Computing Machinery. Komunjer,I.;andNg,S.2011. Dynamicidentificationofdy- namicstochasticgeneralequilibriummodels.Econometrica,

2011

[1] [2]

Theeconomyneedsagent- based modelling.Nature,

Farmer,J.D.;andFoley,D.2009. Theeconomyneedsagent- based modelling.Nature,

2009

[2] [4]

Unveilingthetruth and facilitating change: Towards agent-based large-scale so- cialmovementsimulation

Mou,X.;Wei,Z.;andHuang,X.-J.2024. Unveilingthetruth and facilitating change: Towards agent-based large-scale so- cialmovementsimulation. InFindingsoftheAssociationfor Computational Linguistics. Park, J. S.; O’Brien, J.; Cai, C. J.; Morris, M. R.; Liang, P.; and Bernstein, M. S

2024

[3] [5]

Gao, S.; Wen, Y.; Zhu, M.; Wei, J.; Cheng, Y.; Zhang, Q.; and Shang, S

S3: Social-network simulation sys- tem with large language model-empowered agents.arXiv preprint arXiv:2307.14984. Gao, S.; Wen, Y.; Zhu, M.; Wei, J.; Cheng, Y.; Zhang, Q.; and Shang, S

Pith/arXiv arXiv

[4] [7]

Jin, H.; Yoo, M.; Park, J.; Lee, Y.; Wang, X.; and Kim, J

War and peace (wara- gent): Large language model-based multi-agent simulation of world wars.arXiv preprint arXiv:2311.17227. Jin, H.; Yoo, M.; Park, J.; Lee, Y.; Wang, X.; and Kim, J

Pith/arXiv arXiv

[5] [11]

InProceedings of the Annual ACM Symposium on User Interface Software and Technol- ogy

Generative agents: Interactive simulacra of human behavior. InProceedings of the Annual ACM Symposium on User Interface Software and Technol- ogy. Pearl, J. 2009.Causality: Models, Reasoning and Inference. Cambridge University Press, 2 edition. Piao, J.; Yan, Y.; Zhang, J.; Li, N.; Yan, J.; Lan, X.; Lu, Z.; Zheng,Z.;Wang,J.Y.;Zhou,D.;etal.2025. AgentSocie...

Pith/arXiv arXiv 2009

[6] [12]

Schelling,T.C.1971

Y social: an llm-powered social me- dia digital twin.arXiv preprint arXiv:2408.00818. Schelling,T.C.1971. Dynamicmodelsofsegregation.Jour- nal of mathematical sociology,

Pith/arXiv arXiv 1971

[7] [13]

InProceedings of the Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (System Demonstrations)

Gensim: A general social simulation platform with large language model based agents. InProceedings of the Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (System Demonstrations). Tang,X.;Zou,A.;Zhang,Z.;Li,Z.;Zhao,Y.;Zhang,X.;Co- han,A.;andGerstein,M.2024. MedAgents:Largelangu...

2024

[8] [14]

Touzel,M.P.;Sarangi,S.;Welch,A.;K,G.;Zhao,D.;Yang, Z.;Yu,H.;Gibbs,T.;Kosak-Hine,E.;Musulan,A.;Thibault, C.; Rabbany, R.; Godbout, J.-F.; and Pelrine, K

Simulating social media using large language models to evaluate alternative news feed algorithms.arXiv preprint arXiv:2310.05984, abs/2310.05984. Touzel,M.P.;Sarangi,S.;Welch,A.;K,G.;Zhao,D.;Yang, Z.;Yu,H.;Gibbs,T.;Kosak-Hine,E.;Musulan,A.;Thibault, C.; Rabbany, R.; Godbout, J.-F.; and Pelrine, K

Pith/arXiv arXiv

[9] [15]

Xu, R.; Lu, D.; Tan, X.; Wang, X.; Yuan, S.; Chen, J.; Chu, W.; and Xu, Y

Simulating public administration crisis: A novel generative agent-based simu- lation system to lower technology barriers in social science research.arXiv preprint arXiv:2311.06957. Xu, R.; Lu, D.; Tan, X.; Wang, X.; Yuan, S.; Chen, J.; Chu, W.; and Xu, Y

Pith/arXiv arXiv

[10] [16]

Yang, C.; Zhang, Y.; Wang, J.; He, L.; and Wu, H

Mindecho: Role-playing lan- guage agents for key opinion leaders.arXiv preprint arXiv:2407.05305. Yang, C.; Zhang, Y.; Wang, J.; He, L.; and Wu, H. 2025a. Adeepreinforcementlearningbasedmulti-agentsimulation optimization approach for IGV bidirectional task allocation and charging joint scheduling in automated container termi- nals.Computers & Operations R...

Pith/arXiv arXiv

[11] [17]

Zhang, X.; Lin, J.; Mou, X.; Yang, S.; Liu, X.; Sun, L.; Lyu, H.; Yang, Y.; Qi, W.; Chen, Y.; et al

OASIS: Open agents social interaction simulations on one million agents.arXiv preprint arXiv:2411.11581. Zhang, X.; Lin, J.; Mou, X.; Yang, S.; Liu, X.; Sun, L.; Lyu, H.; Yang, Y.; Qi, W.; Chen, Y.; et al

Pith/arXiv arXiv

[12] [18]

Zhang, X.; Lin, J.; Sun, L.; Qi, W.; Yang, Y.; Chen, Y.; Lyu, H.; Mou, X.; Chen, S.; Luo, J.; et al

Socioverse: A world model for social simulation powered by llm agents and a pool of 10 million real-world users.arXiv preprint arXiv:2504.10157. Zhang, X.; Lin, J.; Sun, L.; Qi, W.; Yang, Y.; Chen, Y.; Lyu, H.; Mou, X.; Chen, S.; Luo, J.; et al

Pith/arXiv arXiv

[13] [19]

Zhao,K.I.;Naim,M.;Kondic,J.;Cortes,M.E.;Ge,J.;Luo, S.; Yang, G

Electionsim: Massive population election simulation powered by large language model driven agents.CoRR. Zhao,K.I.;Naim,M.;Kondic,J.;Cortes,M.E.;Ge,J.;Luo, S.; Yang, G. R.; and Ahn, A. 2024a. LyfeAgents: genera- tive agents for low-cost real-time social interactions.arXiv preprint arXiv:2310.02172. Zhao, Q.; Wang, J.; Zhang, Y.; Jin, Y.; Zhu, K.; Chen, H.;...

Pith/arXiv arXiv

[14] [20]

AlphaAgents: Large Language Model basedMulti-AgentsforEquityPortfolioConstructions.arXiv preprint arXiv:2508.11152, abs/2508.11152. A Main Experiment Details A.1 Parameter Development and Selection We conducted non-exhaustive pilot tests rather than opti- mizing parameters against the reported evaluation metrics. Finalsettingswereselectedbyjointlyconsider...

Pith/arXiv arXiv

[15] [21]

canal maintenance workers,

Responseprobabilitywasselectedjointlywithpopulation size because the expected number of agent responses, and consequently the number of LLM calls, increases with both quantities. For the historical simulations with2,000agents, Parameter Values Examined Final Setting Selection Criterion Response probability0.05,0.2,0.3,0.4,0.5,0.8 0.05for the historical si...

2024

[16] [22]

Schellingparadox

Trading volume was volatile with notable spikes, and the bid-ask spread narrowed over time. These results demonstrate that LLM-driven agents can en- dogenously generate asset bubbles through speculative trad- ing and decision feedback. /uni00000013/uni00000018/uni00000014/uni00000013/uni00000014/uni00000018/uni00000015/uni00000013/uni00000015/uni00000018/...

1971

[17] [23]

Figure 18: Dashboard for managing generated simulation projects

This interface provides a centralized entry point for inspecting project descriptions and initiating subsequent operations. Figure 18: Dashboard for managing generated simulation projects. During simulation execution, the monitoring interface in Figure 19 presents live charts of key metrics together with a dynamic map of agent locations and states. This s...

2023

[18] [113]

G.; and Golub, B.2024

Banerjee, A.; Breza, E.; Chandrasekhar, A. G.; and Golub, B.2024. WhenLessIsMore:ExperimentalEvidenceonIn- formation Delivery During India’s Demonetisation.Review of Economic Studies,

2024

[19] [645]

PrincetonUniversityPress

Hayashi,F.2011.Econometrics. PrincetonUniversityPress. Hu,G.;He,W.;Sun,C.;Zhu,H.;Li,K.;andJiang,L.2023. Hierarchical belief rule-based model for imbalanced multi- classification.Expert Systems with Applications,

2011

[20] [2022]

Rebel on the canal: Disrupted tradeaccessandsocialconflictinChina,1650–1911.Amer- ican Economic Review,

1911

[21] [2023]

Allen, R

GPT-4 Technical Report.arXiv preprint arXiv:2303.08774. Allen, R. C.; Bertazzini, M. C.; and Heldring, L

Pith/arXiv arXiv

[22] [2024]

Li,C.;Su,X.;Han,H.;Xue,C.;Zheng,C.;andFan,C.2023

Episodic Future Thinking MechanismforMulti-agentReinforcementLearning.InPro- ceedings of the Neural Information Processing Systems. Li,C.;Su,X.;Han,H.;Xue,C.;Zheng,C.;andFan,C.2023. Quantifying the impact of large language models on collec- tive opinion dynamics.arXiv preprint arXiv:2308.03313. Li,J.;Lai,Y.;Li,W.;Ren,J.;Zhang,M.;Kang,X.;Wang,S.; Li,P.;Zha...

Pith/arXiv arXiv 2023

[23] [2025]

InPro- ceedings of the CHI Conference on Human Factors in Com- puting Systems

TeachTune: Reviewing Pedagogical Agents Against Diverse Student Profiles with Simulated Students. InPro- ceedings of the CHI Conference on Human Factors in Com- puting Systems. Association for Computing Machinery. Komunjer,I.;andNg,S.2011. Dynamicidentificationofdy- namicstochasticgeneralequilibriummodels.Econometrica,

2011