EconAgentic in DePIN Markets: A Large Language Model Approach to the Sharing Economy of Decentralized Physical Infrastructure

Mocca Schweitzer; Yulin Liu

REVIEW 4 major objections 5 minor 21 references

EconAgentic in DePIN Markets: A Large Language Model Approach to the Sharing Economy of Decentralized Physical Infrastructure

T0 review · 4 major / 5 minor · reviewed 2026-08-05 · deepseek-v4-flash

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

Pith's one-line read Patient autonomous decisions improve stability and inclusion in DePIN token markets without sacrificing efficiency, EconAgentic simulations show.

desk verdict The rejection is right: the paper's central result that higher LLM patience improves DePIN inclusion and stability is largely built into the model's definitions, not an empirical finding about LLM behavior. read the letter →

arxiv 2508.21368 v1 pith:KZBZWXFA submitted 2025-08-29 econ.GN cs.AIq-fin.EC

Yulin Liu , Mocca Schweitzer This is my paper

classification econ.GNcs.AIq-fin.EC

keywords DecentralizedPhysicalInfrastructureLargeLanguageModelsAutonomousAIAgentsMulti-agentSystemsToken-basedEconomicsDePINMarketSimulationStabilityEconomicIncentives

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

EconAgentic is a proposed simulation framework for Decentralized Physical Infrastructure (DePIN) markets — token-incentivized networks of hardware such as wireless hotspots, storage, and GPU compute. The paper models a DePIN project from fundraising through mass adoption, with node providers deciding when to enter and exit based on profitability. The central claim is that when those decisions are made by a large language model, the model's patience — the number of consecutive bad months it tolerates before exiting — reshapes market outcomes: more patient agents raise inclusion, lower token-price volatility, and still allow the market's total capitalization to grow. The authors present this as a step toward designing and governing autonomous, token-based economies so their outcomes stay aligned with efficiency, inclusion, and stability.

What carries the argument

The load-bearing mechanism is the patience parameter inside the LLM-based node-provider agent: the number of consecutive monthly signals (revenue below cost) that must accumulate before the agent answers 'exit'. This single scalar converts a brittle heuristic rule — exit as soon as revenue dips below cost — into a graduated risk tolerance. It is embedded in a closed-form tokenomics model: user count is a scaled count of possible node pairs, revenue is token issuance plus a scaled user term, and token price is total growth-capital endowment divided by tokens on sale. The simulations show that this patience parameter, not the specific LLM prompt or token allocation, is what drives the inclusio

What would settle it

Re-run the same simulation after fitting the user-growth and revenue equations to actual monthly node counts, transaction revenues, and token prices from two existing DePIN projects (e.g., Helium and Filecoin). If higher patience no longer raises inclusion and stability while preserving market-cap growth under the calibrated model, the central claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central claim is that autonomous agents in token-based infrastructure markets serve the whole network better when they are slow to quit. It constructs a closed-form DePIN token economy — user count, node revenue, and token price all follow simple equations — and lets node providers make entry and exit decisions through a small language model; a patience parameter sets how many consecutive unprofitable months must pass before a node exits. The simulations show a consistent direction: more patience raises inclusion and reduces token-price volatility, while market capitalization keeps growing. The stated conclusion is that patient LLM agents mitigate short-term volatility and make d

Load-bearing premise

The paper assumes that a DePIN network's user count and revenue follow two simple equations with made-up constants (100 and a scaling factor k) that are never calibrated to real-world data; if real DePIN revenue behaves differently, the node entry-and-exit dynamics and all reported outcomes change.

Editorial extensions

If this is right

DePIN designers can use this simulator to stress-test token vesting schedules, node costs, and reward rates before deploying smart contracts.
LLM agents with tunable patience can serve as benchmarks for autonomous decision-making in token economies, giving governance a single dial to turn.
Retaining nodes through short-term volatility is a design principle that improves both inclusion and stability without sacrificing measured efficiency.
The efficiency-inclusion-stability triad provides a common yardstick for comparing DePIN designs and AI governance strategies.
The framework can be extended to assess specific token mechanisms — staking, multi-token systems, reward structures — against fairness and sustainability goals.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

Editorial extension: the absolute numbers from the simulation (prices, market caps) should not be read as forecasts, since the revenue and user-growth equations are uncalibrated; the robust content is the ordinal comparison across patience levels.
Editorial extension: the patience mechanism could be applied to the investor side — e.g., growth capitalists who ride out drawdowns instead of selling tokens — which the current model treats only as fixed lifespans.
Editorial extension: calibrating the user-growth and revenue equations to real node counts and transaction data from existing DePIN networks would test whether the patience effect survives contact with empirical parameters.
Editorial extension: the geospatial node-distribution data discussed in the appendix could turn the model spatial, letting patience interact with regional demand shocks rather than global averages.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

The rejection is right, but let me give my own read. The genuinely new thing here is applying LLM-driven agents to a DePIN tokenomics simulation. That combination hasn't been done, and the simulation is simple enough to follow. The three metrics—efficiency, inclusion, stability—are reasonable, and the setup sections are clearly written. The appendix also includes real geospatial data from Helium and the Internet Computer, though it's not used in the simulation.

The problems are where the reader says they are. The headline result—higher patience improves inclusion and stability—is built into the definitions. Inclusion is the fraction of external nodes; if you delay exits for p periods, you mechanically keep more nodes in the network, raising the fraction. Stability is the volatility of log returns; smoothing the exit process reduces supply shocks and hence volatility. That's not LLM-specific. The heuristic benchmark exits immediately when revenue falls below a threshold, which is patience=1, so comparing it to LLM agents with patience>1 biases the comparison. The paper never reports the exact patience values or runs a heuristic with matched patience. So Section 4.2's claim that higher-patience LLMs 'promote decisions that align with long-term trends' isn't supported.

The other soft spots are just as load-bearing. The revenue and user-growth equations in Section 3.2.2 are arbitrary—U(t)=100*sqrt(n(n-1)/2), R_global=P*T_node/n+k*U—with no calibration to real DePIN data. The constants k and 100 are listed as tunable, but there's no sensitivity analysis. Figure 5 has no error bars or confidence intervals. No code is provided. The appendix's heatmaps, while nice, don't validate the model.

I don't think the authors are dishonest; the framework description is honest about its scope. But the central claim is an artifact, and the empirics aren't there. I'd desk reject this in its current form. If the authors reframed the paper as a demonstration of a simulation framework—rather than a substantive finding about LLM behavior—and added matched-patience heuristics plus sensitivity analysis, it could be worth a look. As is, not for peer review.

Referee Report

4 major / 5 minor

Summary. The paper introduces EconAgentic, an agent-based simulation framework for Decentralized Physical Infrastructure (DePIN) markets. It models token distribution with vesting schedules, node-provider entry/exit based on profitability, growth capitalists with stochastic lifespans, and token price as the ratio of growth-capital endowment to tokens on sale. A central design element is a 'patience' parameter that sets the number of consecutive weak signals required before a node exits. The authors compare a threshold heuristic against an LLM-powered agent (gpt-neo-125M) and claim that higher patience improves inclusion and stability without sacrificing efficiency. The paper also lists top DePIN tokens and includes geospatial heatmap visualizations from the Internet Computer dashboard.

Significance. If the framework were validated, it could provide a sandbox for exploring tokenomics and autonomous-agent governance in DePIN markets. The paper has some strengths: it gives explicit mathematical definitions of efficiency, inclusion, and stability; it uses an identifiable open-source LLM checkpoint; it includes a real market-cap table; and it makes its simulation mechanics transparent enough for parameters to be audited. However, the headline result is not established. The patience effect is a mechanical consequence of the exit-rule definition, the model equations are uncalibrated, and the simulation lacks error bars, sensitivity analysis, and reproducibility details. As it stands, the paper reads as a prototype/position piece rather than a confirmatory study; its contribution to the DePIN economics literature is limited unless the central claims are re-derived with matched controls and empirical calibration.

major comments (4)

[Section 4.2, Figure 5] The headline claim that 'LLMs with higher patience are more effective' is forced by construction. Patience is defined as the number of consecutive weak signals required before a node exits. Inclusion is I_inc=(N_total-N_init)/N_total, so delaying exits keeps N_total higher and mechanically increases inclusion. Stability is the standard deviation of log returns, so smoothing the exit process reduces token-supply volatility by construction. The heuristic benchmark in Section 3.3.1 exits whenever R_global<tau*C_node, i.e., patience=1. The paper neither reports the patience values used for the LLM nor provides a heuristic run with matched patience. Figure 5 therefore cannot distinguish LLM-specific reasoning from the low-pass-filter effect of the patience parameter. This is load-bearing because the abstract and Section 4.2 attribute the outcome to the LLM.
[Section 3.2.2] The revenue model is arbitrary and uncalibrated. U(t)=100*sqrt(n(n-1)/2) and R_global(t)=P(t-1)*T_node(t)/n(t-1)+k*U(t) contain unspecified constants 100 and k. Node entry and exit depend entirely on R_global relative to C_node, so all downstream metrics inherit this arbitrariness. No values are given for k, C_node, tau_node, or the log-normal parameters of GC lifespans, and no fitting is done to any of the DePIN projects in Table 2. The paper itself concedes in Section 5.3 that empirical data could be added. Without calibration and sensitivity analysis, the simulation cannot support claims about DePIN markets.
[Section 4.2] The reported results are single-trajectory qualitative curves. Figure 5 shows monotonic improvement with patience but gives no error bars, no multiple seeds, no confidence intervals, and no sensitivity to the arbitrary parameters. Because the simulation is stochastic (log-normal GC lifespans, LLM response variability), one run cannot establish that the ordering is stable. Multiple independent runs with variance reporting and a formal comparison across patience levels are needed before the stated conclusions can be drawn.
[Section 3.3.2 and 4.2] The LLM implementation is not specified in a reproducible way. The prompts shown are generic; the paper does not state how 'patience' is encoded for the LLM, how many LLM calls are made per period, what temperature and parsing rules are used, or how many simulation runs are averaged. The heuristic baseline is not matched to the LLM in terms of patience. This omission prevents verification that the difference in Figure 5 is due to LLM reasoning rather than prompt wording or decision-rule details. Given that the central comparison is LLM-based versus heuristic-based agents, this is a load-bearing reproducibility gap.

minor comments (5)

[Appendix A] Typos: 'epicts' should be 'depicts' (Figure 8); 'Coutnries' should be 'Countries' (Figure 9); Section 3.1 'So on and so forth' should be 'and so on'.
[Section 3.2.3] 'the r tokens are sold' should read 'their tokens are sold'.
[Section 2.2.1] The '$100,000 per node' assumption conflicts with the earlier range of $500-$20,000 for real DePIN hardware. It should be labeled as an illustrative simplification with a sensitivity discussion.
[Section 5.3] The paper uses the term 'benchmarks' but does not release a benchmark dataset or task suite. Consider renaming to 'evaluation framework' or providing the code and data to make the benchmark concrete.
[Abstract and Section 4.2] 'Human heuristic benchmarks' is misleading: the baseline is a simple threshold rule, not human behavior. Suggest 'rule-based benchmark'.

Circularity Check

2 steps flagged · score 8.0 of 10

Patience result is definitional: higher patience mechanically raises inclusion and smooths exit-driven volatility; the LLM comparison does not validate the central claim.

self definitional [Section 4.2 (Figure 5) and Section 4.1.2 (I_inc definition)]
"A key variable in this model is the patience parameter, which governs the number of consecutive signals required for a node to exit the system, thus representing varying levels of risk tolerance. ... Inclusion increases with higher patience, as nodes are retained in the system longer, allowing for more favorable market conditions to emerge. This results in a more inclusive network with fewer premature exits."

Inclusion is defined as I_inc = (N_total - N_init)/N_total, a monotonically increasing function of N_total for fixed N_init. Patience is defined as requiring more consecutive weak signals before a node exits, which mechanically keeps nodes in N_total longer. Therefore 'higher patience -> higher inclusion' is true by construction for any agent, heuristic or LLM, independent of reasoning quality. The paper's own explanation concedes the mechanism is retention, not LLM judgment.
self definitional [Section 4.2 versus Section 3.3.1]
"Similarly, stability improves with greater patience, as nodes become less sensitive to short-term market fluctuations, leading to reduced volatility and smoother system behavior. ... Node Exit: A node exits if R_global(t) < tau_node×C_node, where tau_node is the node's risk tolerance."

Stability is defined as the standard deviation of log returns of token price. The patience parameter is a minimum count of consecutive threshold violations before exit, i.e. a low-pass filter on the exit signal. The heuristic benchmark exits on the first violation (patience=1), so the comparison tests the mechanical smoothing effect of delayed exits, not LLM-specific decision-making. The LLM does not choose or learn patience; it is an input variable. The reported improvements in stability follow from the exit-rule parameterization, not from language-model reasoning.

full rationale

The central quantitative results of Section 4.2 reduce to the definition of the patience parameter. Inclusion is monotonically increasing in total nodes, and patience is defined as retaining nodes longer, so higher patience mechanically raises inclusion. Stability is a volatility measure, and patience temporally smooths the exit signal; the heuristic benchmark with immediate exit is equivalent to patience=1, making the comparison a test of a smoothing parameter rather than LLM cognition. The paper itself states that the mechanism is retention and reduced sensitivity to fluctuations. The revenue model and uncalibrated constants (100, k) in Section 3.2.2 are an external-validity and calibration risk, not circularity, because outcomes are not defined in terms of those constants. The same-author citation [14] appears only in related-work/future-directions and is not load-bearing for the simulation. Accordingly, the main claim is forced by construction, yielding a score of 8.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The entire simulation rests on arbitrary, uncalibrated functional forms and parameters. The ledger is heavy with ad hoc assumptions, and the reported findings are direct consequences of these choices.

free parameters (6)

k
Scaling factor for user-related revenue in R_global(t); value never given, arbitrary.
User growth multiplier (100) = 100
Constant multiplying sqrt(n(n-1)/2) in U(t); chosen without empirical basis.
C_node
Node operating cost per month; no value specified, yet determines entry and exit.
Patience (exit threshold) = varied
Number of consecutive weak signals before exit; swept across simulations, and the main driver of the reported results.
tau_node (heuristic risk tolerance)
Threshold for heuristic node exit; not specified.
GC lifespan log-normal parameters
Mean and variance of the log-normal distribution for growth capitalist lifespans; unspecified.

assumptions (5)

domain assumption Metcalfe's Law: network value V proportional to n^2
Used to justify network growth and the user formula in Section 2.2.1.
ad hoc to paper User count follows U(t) = 100 * sqrt(n(t)*(n(t)-1)/2)
Arbitrary functional form for the number of users as a function of nodes, stated without derivation or empirical support in Section 3.2.2.
ad hoc to paper Revenue model: R_global(t) = P(t-1)*T_node(t)/n(t-1) + k*U(t)
Determines node profitability and hence entry and exit; no theoretical or empirical justification given.
ad hoc to paper Token price is total GC endowment divided by tokens on sale
Simple price formation rule in Section 3.2.4; ignores order books, liquidity, and speculation.
domain assumption LLM model responses (gpt-neo-125M) reflect rational economic decisions
The entire comparison rests on treating a 125M-parameter LM's yes or no answers as meaningful economic choices, with no validation against human behavior.

how reviews work

0 comments

Cite this review

Pith. "Pith review of EconAgentic in DePIN Markets: A Large Language Model Approach to the Sharing Economy of Decentralized Physical Infrastructure." pith.science (2026). https://pith.science/paper/KZBZWXFA

@misc{pith2026250821368,
  author       = {Pith},
  title        = {Pith review of: EconAgentic in DePIN Markets: A Large Language Model Approach to the Sharing Economy of Decentralized Physical Infrastructure},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/KZBZWXFA}},
  note         = {Machine review of arXiv:2508.21368}
}

abstract

The Decentralized Physical Infrastructure (DePIN) market is revolutionizing the sharing economy through token-based economics and smart contracts that govern decentralized operations. By 2024, DePIN projects have exceeded \$10 billion in market capitalization, underscoring their rapid growth. However, the unregulated nature of these markets, coupled with the autonomous deployment of AI agents in smart contracts, introduces risks such as inefficiencies and potential misalignment with human values. To address these concerns, we introduce EconAgentic, a Large Language Model (LLM)-powered framework designed to mitigate these challenges. Our research focuses on three key areas: 1) modeling the dynamic evolution of DePIN markets, 2) evaluating stakeholders' actions and their economic impacts, and 3) analyzing macroeconomic indicators to align market outcomes with societal goals. Through EconAgentic, we simulate how AI agents respond to token incentives, invest in infrastructure, and adapt to market conditions, comparing AI-driven decisions with human heuristic benchmarks. Our results show that EconAgentic provides valuable insights into the efficiency, inclusion, and stability of DePIN markets, contributing to both academic understanding and practical improvements in the design and governance of decentralized, tokenized economies.

Figures

Figures reproduced from arXiv: 2508.21368 by the authors.

**Figure 1.** Contributions. To bridge these gaps, we introduce EconAgentic, a comprehensive framework designed to empower both academic researchers and industry practitioners to understand, analyze, and design DePIN markets in ways that ensure their outcomes align with human values. Our framework is both grounded in industry practices and capable of abstracting scientific inquiry at the micro and macro levels. As in [PITH_FULL_… view at source ↗

**Figure 2.** CePIN vs. DePIN. economy models, like Uber and Airbnb, by decentralizing the platforms themselves. In DePIN ecosystems, networks of IoT-enabled devices facilitate real-time data exchange and operational decisions, making them smarter and more interconnected [9, 18]. As shown in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗

**Figure 3.** DePIN Market Evolution Dynamics Horowitz in 20218 , while Render Network raised $30 million from Multicoin Capital9 . For simplicity, we assume the project raises $10 million in a single venture capital round. Half of the raised funds ($5 million) are allocated to development and operations, while the other $5 million is used to deploy the initial network nodes. Setup costs can vary significantly, ranging from $500 … view at source ↗

Figures from the paper (8 more)

**Figure 4.** Figure 4: Token distribution schedules for the core team, VCs, [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

**Figure 5.** Figure 5: Comparison of Efficiency, Inclusion, and Stability Across Models. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: Datacenter Density Heatmap. The Internet Computer protocol, developed by Dfinity Foundation, offers the detailed geospatial data of data center locations, along with the number of active nodes and node providers in each data center, available at the Internet Computer …

**Figure 7.** Figure 7: Data Centers by Continent [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]

**Figure 8.** Figure 8: Node Machines by Continent Another notable platform is DePINscan, which aggregates data from sources like the W3bstream Trusted Metrics API and thirdparty APIs to provide a comprehensive overview of DePIN projects. Developers can submit their projects for review, and …

**Figure 10.** Figure 10: The Diluted Market Cap [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]

**Figure 11.** Figure 11: The Number of Nodes [PITH_FULL_IMAGE:figures/full_fig_p011_11.png]

**Figure 12.** Figure 12: The Number of Users [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

21 extracted references · 14 canonical work pages

[1]

Daron Acemoglu. 2021. Harms of AI. Oxford Handbook of AI Governance (2021). https://www.nber.org/papers/w29247 Forthcoming

work page 2021
[2]

Daron Acemoglu and Pascual Restrepo. 2018. The race between man and machine: Implications of technology for growth, factor shares, and employment. American Economic Review 108, 6 (2018), 1488–1542. https://doi.org/10.1257/aer.20170672

work page doi:10.1257/aer.20170672 2018
[3]

Shuchih E Chang and Yichian Chen. 2020. When blockchain meets supply chain: A systematic literature review on current development and potential applications. Ieee Access 8 (2020), 62478–62494

work page 2020
[4]

Michael T. C. Chiu, Sachit Mahajan, Mark C. Ballandies, and Uroš V. Kal- abić. 2024. DePIN: A Framework for Token-Incentivized Participatory Sensing. arXiv:2405.16495 [cs.GT]

work page Pith review arXiv 2024
[5]

Cointelegraph. 2023. Decentralized Physical Infrastructure Network (De- PIN), explained. https://cointelegraph.com/explained/decentralized-physical- infrastructure-network-depin-explained. https://cointelegraph.com/explained/ decentralized-physical-infrastructure-network-depin-explained Accessed: 2024- 05-28

work page 2023
[6]

Han Ding, Yinheng Li, Junhao Wang, and Hang Chen. 2024. Large Language Model Agent in Financial Trading: A Survey. arXiv preprint arXiv:2408.06361 (2024)

arXiv 2024
[7]

Xinxin Fan and Lei Xu. 2023. Towards a Rollup-Centric Scalable Architecture for Decentralized Physical Infrastructure Networks: A Position Paper. In Proceedings of the Fifth ACM International Workshop on Blockchain-enabled Networked Sensor Systems. 9–12

work page 2023
[8]

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. Measuring massive multitask language under- standing. arXiv preprint arXiv:2009.03300 (2020)

arXiv 2020

Show all 21 references

[9]

Steve Huckle, Rituparna Bhattacharya, Martin White, and Natalia Beloff. 2016. In- ternet of things, blockchain and shared economy applications. Procedia computer science 98 (2016), 461–466

2016
[10]

IoTeX IoTeX. 2018. a Decentralized Network for Internet of Things Powered by a Privacy-Centric Blockchain. White Paper 1 (2018)

2018
[11]

Tintin Land. 2021. Before Scaling to Mass Adoption: How does DePin Break Through Technology, Data, and Market? https: //medium.com/@tintin.land2021/before-scaling-to-mass-adoption-how- does-depin-break-through-technology-data-and-market-72f79243f76c. https://medium.com/@tintin...

2021
[12]

Nian Li, Chen Gao, Mingyu Li, Yong Li, and Qingmin Liao. 2024. Econagent: large language model-empowered agents for simulating macroeconomic activities. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) . 15523–15536

2024
[13]

Xiao Liu, Hao Yu, Hanchen Zhang, Yifan Xu, Xuanyu Lei, Hanyu Lai, Yu Gu, Hangliang Ding, Kaiwen Men, Kejuan Yang, et al. 2023. Agentbench: Evaluating llms as agents. arXiv preprint arXiv:2308.03688 (2023)

2023 arXiv
[14]

Yulin Liu and Luyao Zhang. 2024. The Economics of Blockchain Governance: Eval- uate Liquid Democracy on the Internet Computer.arXiv preprint arXiv:2404.13768 (2024)

2024 arXiv
[15]

Katya Malinova and Andreas Park. 2023. Tokenomics: when tokens beat equity. Management Science 69, 11 (2023), 6568–6583

2023
[16]

Shaoguang Mao, Yuzhe Cai, Yan Xia, Wenshan Wu, Xun Wang, Fengyi Wang, Tao Ge, and Furu Wei. 2023. Alympics: Language agents meet game theory.arXiv preprint arXiv:2311.03220 (2023)

2023 arXiv
[17]

Ahmed Afif Monrat, Olov Schelén, and Karl Andersson. 2019. A survey of blockchain from the perspectives of applications, challenges, and opportunities. Ieee Access 7 (2019), 117134–117151

2019
[18]

Rodrigo Roman, Jianying Zhou, and Javier Lopez. 2013. On the features and challenges of security and privacy in distributed internet of things. Computer networks 57, 10 (2013), 2266–2279

2013
[19]

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, et al. 2022. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv prep...

2022 arXiv
[20]

Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, et al. 2023. The rise and potential of large language model based agents: A survey. arXiv preprint arXiv:2309.07864 (2023)

2023 arXiv
[21]

Luyao Zhang. 2023. The Future of Finance: Synthesizing CeFi and DeFi for the Benefit of All. In Financial Literacy in Today´s Global Market, Ireneusz Miciuła (Ed.). IntechOpen, Rijeka, Chapter 13. https://doi.org/10.5772/intechopen.1003042 A THE DATA DePIN projects exhibit dis...

2023 doi

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.

[1] [1]

Daron Acemoglu. 2021. Harms of AI. Oxford Handbook of AI Governance (2021). https://www.nber.org/papers/w29247 Forthcoming

work page 2021

[2] [2]

Daron Acemoglu and Pascual Restrepo. 2018. The race between man and machine: Implications of technology for growth, factor shares, and employment. American Economic Review 108, 6 (2018), 1488–1542. https://doi.org/10.1257/aer.20170672

work page doi:10.1257/aer.20170672 2018

[3] [3]

Shuchih E Chang and Yichian Chen. 2020. When blockchain meets supply chain: A systematic literature review on current development and potential applications. Ieee Access 8 (2020), 62478–62494

work page 2020

[4] [4]

Michael T. C. Chiu, Sachit Mahajan, Mark C. Ballandies, and Uroš V. Kal- abić. 2024. DePIN: A Framework for Token-Incentivized Participatory Sensing. arXiv:2405.16495 [cs.GT]

work page Pith review arXiv 2024

[5] [5]

Cointelegraph. 2023. Decentralized Physical Infrastructure Network (De- PIN), explained. https://cointelegraph.com/explained/decentralized-physical- infrastructure-network-depin-explained. https://cointelegraph.com/explained/ decentralized-physical-infrastructure-network-depin-explained Accessed: 2024- 05-28

work page 2023

[6] [6]

Han Ding, Yinheng Li, Junhao Wang, and Hang Chen. 2024. Large Language Model Agent in Financial Trading: A Survey. arXiv preprint arXiv:2408.06361 (2024)

arXiv 2024

[7] [7]

Xinxin Fan and Lei Xu. 2023. Towards a Rollup-Centric Scalable Architecture for Decentralized Physical Infrastructure Networks: A Position Paper. In Proceedings of the Fifth ACM International Workshop on Blockchain-enabled Networked Sensor Systems. 9–12

work page 2023

[8] [8]

Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2020. Measuring massive multitask language under- standing. arXiv preprint arXiv:2009.03300 (2020)

arXiv 2020