Axient: On-Chain Credit and Loss Allocation for Leveraged Event Markets: A Venue-Agnostic Protocol for Traders, Credit Providers, Market Makers, and Liquidation Backstops

Maksym Nechepurenko

REVIEW 4 major objections 6 minor 18 references

Axient: On-Chain Credit and Loss Allocation for Leveraged Event Markets: A Venue-Agnostic Protocol for Traders, Credit Providers, Market Makers, and Liquidation Backstops

T0 review · 4 major / 6 minor · reviewed 2026-08-04 · deepseek-v4-flash

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

Pith's one-line read This paper argues that leveraged event-market credit can be structured on-chain with deterministic loss allocation, and that a synthetic agent stress test shows layered protection reduces but does not eliminate Senior losses, while addition

desk verdict A carefully hedged, artifact-rich protocol-design paper whose formal accounting layer is solid and whose agent-based findings are honestly labeled synthetic; the main open risk is the inherited position-level certificate (Assumption 3.9), so it deserves refereeing but not trust as a production blueprint. read the letter →

arxiv 2608.00647 v1 pith:B5FSO5WX submitted 2026-08-01 q-fin.TR q-fin.CPq-fin.RM

Maksym Nechepurenko This is my paper · ORCID

classification q-fin.TRq-fin.CPq-fin.RM

keywords decentralizedfinancepredictionmarketseventcontractsagent-basedcomputationaleconomicsmarginliquidationcreditpoolslosswaterfalls

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

The paper develops a venue-agnostic on-chain architecture for financing leveraged event positions, separating traders, Senior Credit LPs, market makers, liquidators, and Liquidation Backstop Providers. It formalizes accounting identities, a deterministic loss waterfall, withdrawal queues, and an endogenous capital market, then stress-tests the design with a fixed-seed agent-based model. The central claim is that the protocol's contract guarantees (like balanced accounting and junior-before-senior loss allocation) hold exactly, while the economic layer's behavior depends on author-specified response rules. The agent experiments show layered protection cuts two-year Senior-loss incidence from 99.0% to 57.3% but still admits only 38% of requested credit, impairs LBP capital in 94.8% of paths, and reveals that bonded market-maker capacity can increase aggregate shortfall. A sympathetic reader should care because the paper demonstrates a method for separating enforceable protocol guarantees from unvalidated assumptions about participant behavior.

What carries the argument

The central mechanism is a state-transition system with a deterministic loss waterfall (ordered capacities from position buffers through LBP capital to Senior principal) combined with a utilization-based rate curve and an endogenous capital-market fixed point. Named results include the ordered waterfall operator, the utilization fixed point theorem, and the capacity-induced-shortfall condition, which together separate contract-enforceable accounting from behavioral response functions governing provider participation, trader leverage demand, market-maker delivery, and liquidator entry.

What would settle it

Run the same protocol with empirically calibrated agent behavior from a real prediction-market venue, observing actual LP participation, trader leverage demand, market-maker delivery, and liquidator entry over a comparable horizon; if the realized response functions deviate materially from the assumed forms, the H1-H7 outcomes—particularly the 159.1% leverage-sensitivity ratio and the H4 capacity-induced shortfall—would not be reproduced.

Watch

Extended reading notes

Core claim

The protocol can be decomposed into a deterministic accounting core and a behavioral economic layer. The core guarantees balanced real- and integer-unit accounting, settlement-confirmed debt priority, trader residual ownership, idempotent partial settlement, a unique and conservative loss waterfall, and request-time-neutral loss-participating withdrawal queues. The behavioral layer, modeled with heterogeneous agents, yields six directional findings: layered protection reduces Senior-loss incidence but does not eliminate it; higher leverage is a distinct capital regime, with 5x-heavy raising Senior-loss incidence by 159.1% versus 2x-only; and the registered hypothesis that bonded market-maker

Load-bearing premise

The agent-based results assume author-specified behavioral response functions and shock distributions, so if real providers, traders, or venues respond differently, the quantitative findings (like the 57.3% Senior-loss incidence) do not transfer; the protocol also inherits an unverified assumption about an external certified debt-clearing interface.

Editorial extensions

If this is right

If the architecture is adopted, layered protection can reduce but not eliminate Senior principal loss, so the protocol must publicly expose residual senior impairment risk rather than describe itself as insured.
5x leverage should be treated as a materially different capital regime from 2x, requiring stricter admission, reserve, and execution-capacity gates.
Bonded market-maker capacity should not be credited one-for-one as reserve substitution; incremental admission must be capped below the protected capacity increment to avoid the capacity-induced-shortfall paradox.
Utilization-only interest pricing can be wrong-way under stress, lowering rates while expected loss rises, so a separate risk spread is needed.
Strict pool isolation prevents direct liability transfer but does not eliminate economic contagion from common stablecoin or venue factors, as shown by the 0.842 cross-pool shortfall correlation.

Reading between the lines

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

The failure of the bonded-MM hypothesis suggests a general principle for DeFi risk: any protective mechanism that also expands the admission envelope can be self-defeating unless admission growth is explicitly bounded by the stressed incremental protection it provides.
If the synthetic responses were replaced with empirically calibrated behavior from live venues and LPs, the quantitative findings (57.3% Senior-loss incidence, 38% credit acceptance) would likely change, but the qualitative separation between guaranteed accounting and behavioral uncertainty would likely persist.
The role separation between liquidators (execution) and LBPs (loss-bearing capital) could be applied beyond event markets to other leveraged DeFi products, such as perp margin, where execution capacity and loss absorption are routinely conflated.
A testable extension would re-run the agent model with alternative behavioral rules (e.g., learning agents or different herding parameters) to check whether the H4 failure and the wrong-way pricing result are robust to behavioral misspecification.

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

Worth reading. The formal core—five-role separation, deterministic loss waterfall, loss-participating withdrawal queue, protection contagion, wrong-way utilization pricing, and the capacity-induced-shortfall result (Prop 16.3)—is genuinely new and cleanly stated. The artifact story is unusually strong: 28 exact fixtures, 31,082 deterministic checks, 70M zero-failure invariant evaluations, pinned container, and an honestly reported failed hypothesis (H4). The paper never overclaims; it repeatedly labels outputs as synthetic mechanism comparisons and discloses the founder's financial interest. That discipline earns real credit.

The soft spots are real but mostly conditionable. The agent layer is author-specified and uncalibrated; H1–H7 cannot transfer to any observed market. The paper says so itself, so I treat that as a boundary, not a flaw, though the headline numbers (99%→57.3%, 94.8% LBP impairment) will be misread unless reviewers and editors emphasize the caveat. The deeper concern is Assumption 3.9: the certified debt-clearing interface is inherited by citation from an inaccessible prior paper, and it carries the venue-agnostic architecture. For the non-native backing modes admitted by the taxonomy, the lower settlement envelope and upper cost bound depend on venue behavior the protocol cannot enforce. If that primitive fails, the settlement-confirmed guarantees in Sections 4, 8, 10, and 20 do not transfer. The paper is honest that this is an assumption, but the headline claim is exactly what it is asked to carry. Also minor: the reference snapshot is access-controlled and the self-citation density is high, though the cited prior work does appear to be the actual foundation.

My overall take: this is a competent, self-aware protocol-design preprint with a solid formal layer and honestly bounded experiments. The central architecture holds up as a conditional statement. What it needs is peer review focused on Assumption 3.9 and external calibration—not rejection.

I'd send it to a serious referee. The right referee will check the certificate construction, the waterfall proofs, and the simulation's boundary conditions. It should not be desk-rejected.

Referee Report

4 major / 6 minor

Summary. The paper develops a venue-agnostic on-chain credit and loss-allocation protocol for leveraged event markets, separating traders, Senior Credit LPs, market makers, liquidators, and Liquidation Backstop Providers. It formalizes pool and debt accounting, a deterministic loss waterfall, withdrawal queues, collateralized MM commitments, utilization-based pricing, and an endogenous capital-market model with agent-based simulations. Formal results cover balanced accounting, settlement-confirmed debt priority, waterfall existence/uniqueness/conservation, utilization fixed points, and capacity-induced shortfall. The release includes 28 exact fixtures, 31,082 deterministic checks, and a fixed-seed agent experiment with 96 paths and six of seven hypotheses passing; H4 (bonded MM reduces shortfall) fails. The paper explicitly labels all agent-based outputs as synthetic and disclaims production forecasts.

Significance. If the formal core and the 'synthetic mechanism comparison' framing are accepted, this is a substantial reference architecture for a novel type of DeFi credit protocol. Strengths include: extensive machine-checked accounting invariants (31,082 deterministic checks), exact closed-form waterfall and conservation proofs, an honest and detailed limitations section, explicit release-registered but not externally preregistered hypotheses, and a negative result (H4) that is reported rather than tuned away. The separation of contract guarantees, behavioral mechanisms, and external assumptions is a useful discipline. However, the economic contribution is conditional on author-specified behavioral rules and a shock structure with no calibration, and the position-level primitive is inherited from an inaccessible prior paper. The paper's significance is therefore primarily as a formal/simulation reference design, not as empirical evidence about production performance.

major comments (4)

[Assumption 3.9; Sections 4, 8, 20] The central venue-agnostic guarantees ('settlement-confirmed debt priority', 'no collateral escape', waterfall conservation) all rely on the certified debt-clearing interface of Assumption 3.9. The paper neither proves nor restates the construction, citing instead a separate paper and a private repository. For venue-capability modes 2–4 (Definitions 3.6–3.7), the lower settlement envelope B and upper cost bound H are not contract-controlled; if the venue delays settlement or refuses delegated liquidation, inequality (6) can fail, and then the protocol-level guarantees are conditional on an unresolved external assumption. This should be stated explicitly as a limitation at the point of the assumption, with a concrete statement of what must be verified per venue mode before the guarantees transfer.
[Sections 18–19; Eqs. (76)–(79), (95)] The main new economic findings (e.g., Senior-loss incidence reduction from 99.0% to 57.3%, H4 failure, 0.842 cross-pool correlation) are outputs of a fixed-seed simulator whose participation functions, trader utility, withdrawal map, and Markov shock structure are all author-specified. The paper honestly labels these as synthetic, but the abstract and results sections still present them as headline findings. As a referee I do not treat the lack of calibration as an error—the paper disclaims it—but the conditional nature should be more prominent at the point of the headline numbers, and the fixed-point and contraction results (Theorems 15.3 and Proposition 15.4) should be tied to the fact that the economic 'fixed point' is only as good as the assumed response functions.
[Section 19, Tables 13 and 17; H2] H2 ('Full Axient preserves at least 70% of Senior-only accepted credit demand') is reported as passing with 155.6%, but the comparison baseline is the Senior-only configuration which admits only 23.3% of requested credit. The 38.0% admitted by Full Axient is still a rejection of 62% of demand. The paper does note this, but the metric as defined measures relative preservation from a very low baseline; this is a load-bearing interpretation issue for the claim that endogenous supply 'supports more accepted credit'. The text should clarify that this is not evidence of adequate credit supply in absolute terms, and the headline 'H2 passes' should be read with this caveat.
[Section 14.6, Table 6; Eq. (75)] The Monte Carlo table reports only binomial standard errors for incidence, with no error bars for means or return quantiles, and the agent-based tables (Section 19) report no measures of Monte Carlo variability over the 96 paths. The agent model has a fixed seed, which makes point estimates reproducible but does not quantify sampling uncertainty over paths or parameter sensitivity. The paper discusses this limitation qualitatively, but the phrase 'mean synthetic annualized Senior return rises to 20.6%' is reported without any dispersion or standard error. At minimum, the paper should state the standard error of the 96-path incidence estimates, which materially affects how precisely H1's 42.1% relative reduction can be claimed.

minor comments (6)

[Title/Abstract] 'Axient' is introduced without a formal definition; if this is intended as a protocol name this is fine, but the abstract could state explicitly that all results are for the specified transition system and synthetic parameterization, not for a live deployment.
[Section 3.9 and Appendix F] The paper repeatedly refers to a 'private repository' and a 'tagged reference snapshot' with hashes. While the full source export is said to be in the research package, the minimal arXiv package does not include it, so the verification claims are not independently checkable from the submitted manuscript. Consider including the full source export or a public mirror for the formal claims.
[Equation (74)] The synthetic haircut generator includes a T5 term and beta components; a sentence connecting this to venue-specific settlement risk and the position-level certificate of Assumption 3.9 would improve clarity.
[Section 19.4] The phrase 'Aggregate run-week incidence falls from 4.1 percent in Senior only to 2.6 percent in full phased' uses 'run-week incidence' but the empirical definition of a run threshold is only in the appendix; the text should define it at first use.
[Appendix D.10] Table 33's 'Haircut Risk' column lists 0.655x, 0.905x, 1.203x; these are labeled as synthetic but it is unclear whether smaller or larger is riskier. Add an explicit statement.
[Throughout] The distinction between 'reserve' and 'pool reserve' is used consistently, but the figure captions (e.g., Figure 4) would benefit from stating the specific experiment parameters inline (e.g., loan maturity generator) rather than only in the appendix.

Circularity Check

1 steps flagged · score 4.0 of 10

Formal core is self-contained; one load-bearing primitive is imported from the author's own prior paper via Assumption 3.9.

self citation load bearing [Section 3.6, Assumption 3.9; also Section 1 ('The position-level primitive is inherited from Nechepurenko [2026a]')]
"The position-level primitive is inherited from Nechepurenko [2026a]. ... Assumption 3.9 (Certified debt-clearing interface). For each admitted position i, the risk layer publishes C_i = (x̂_i,u, B_i,u,Δ, H_i,u,Δ, m_i, Δ_i) with K_i,u + B_i,u,Δ(x̂_i,u) ≥ H_i,u,Δ + m_i."

Every protocol-level guarantee about settlement-confirmed debt priority, the deterministic loss waterfall, and 'no collateral escape' is conditioned on inequality (6) being satisfiable. This paper neither proves nor reconstructs the certificate; it cites the author's own Nechepurenko [2026a]. The load-bearing premise is therefore a same-author citation rather than an argument developed in this paper. It is not a fitted-value tautology — the paper is transparent that it is an assumption — but the derivation chain for the protocol-level results passes through this self-citation, making it the one substantive circularity-adjacent dependency.

full rationale

The accounting identities, waterfall conservation, queue neutrality, and utilization fixed-point theorems are derived from stated definitions and are self-contained; they do not reduce to fitted values or to hidden equivalences. The agent-based 'findings' are outputs of a simulator whose behavioral rules, shock structure, and hypothesis thresholds are author-specified and explicitly labeled synthetic, not calibrated to observations; because no parameter is fitted to data and the paper repeatedly disclaims production forecasting (Sections 1.3, 19.11, 23), the H1–H7 results are not 'predictions' that reduce by construction. The only load-bearing external dependency is Assumption 3.9, which imports the certified debt-clearing interface from the author's own prior work. This is a genuine self-citation at a central point, but because it is stated as an assumption and the paper's main accounting, waterfall, and endogenous-market contributions have independent derivational content, the overall circularity score is moderate rather than high.

Assumptions & free parameters 8 free parameters · 5 assumptions · 3 invented entities

The paper's formal layer rests on standard math plus two domain assumptions (certified debt-clearing interface, single-asset book). The economic layer rests on a lattice of author-chosen parameters: rate curve, haircut generator, interest splits, protection sizes, agent pool calibrations, regime matrix, reserve rule, and fixed seeds. Every H1-H7 verdict is a function of these choices. No entity introduced (LBP, queue, MM bond) has independent falsifiable evidence outside the paper itself. This is the honest measure of what the paper contributes: a structurally complete but empirically ungrounded mechanism design.

free parameters (8)

Borrow-rate curve parameters (r0, U*, s1, s2) = 0.06, 0.80, 0.12, 0.65
Author-specified utilization pricing policy input, not estimated from market data (§D.2, eq 27).
Synthetic haircut generator coefficients in eq (74) = 0.012, 0.055, 0.018, 0.30, 0.28, 0.35, 0.60, 0.008; tier shifts (0, 0.010, 0.030)
All distributions, incident probabilities, and tier shifts are author-chosen (§D.4).
Interest-allocation weights (alpha_S, alpha_J, alpha_R, alpha_P, alpha_M) = 0.90→0.73 senior; 0→0.13 reserve; 0→0.06 LBP; 0→0.03 MM; 0.10→0.05 protocol
Protection layers are priced by these policy fixtures (§D.5).
Protection layer sizes = reserve 1%, LBP 3% of senior supply; bonded MM 1.5%/4.5% of 3x/5x notional
Capital-layer capacities are transparent policy fixtures (§D.5).
Agent pool parameters = Senior targets 10M/8M/7M; haircuts 1.0%/1.4%/2.0%; risk multipliers 0.655/0.905/1.203; MaxL 5/5/3
Pool-specific synthetic calibration (Table 33, §D.10).
Markov regime matrix, multipliers, shock probabilities = diag 0.955/0.720/0.500; multipliers 1.0/2.5/6.0; per-week shock probabilities listed in §D.10
The entire stress structure is author-specified; H6's 0.842 correlation is built by the common-factor generator.
Dynamic reserve target rule = 2.5x perceived loss, 8pp cap, 60/40 senior/protocol funding split
Reserve controller is a transparent fixture, not an optimal policy (§D.10).
Agent experiment dimensions = 96 paths, 104 weeks, 636 agents/path, PCG64 seed 20260720
Fixed-seed design choices that determine all H1-H7 verdicts (§D.10).

assumptions (5)

domain assumption Assumption 3.9 — every admitted position provides a valid certified debt-clearing interface from the prior position-level mechanism
Inherited by citation from Nechepurenko [2026a]; not proven or independently verified here. Protocol-level admission and the entire loss waterfall assume certificate validity.
domain assumption Assumption 4.1 — single-asset book: all terms denominated in one settlement asset; no heterogeneous stablecoin netting at par
Scopes the accounting closure theorems; cross-asset and bridge risk are excluded by assumption.
ad hoc to paper Provider participation functions G_i, H_j, trader utility U_n,t, and withdrawal map F are continuous, bounded, monotone response rules with author-chosen forms
The fixed-point results (Thm 15.3, Prop 15.4, Prop 17.1) hold for these behavioral classes, not for real agents.
standard math Brouwer fixed-point and Banach contraction arguments for utilization and withdrawal equilibria
Applied within the stated continuity and contraction conditions; no controversy.
ad hoc to paper Markov regime and shock structure (common venue incidents, stablecoin depeg, oracle delay, strategic MM withdrawal) with fixed transition probabilities
The common-factor contagion result H6 is a property of this generator, not of production markets; the paper admits this.

invented entities (3)

Liquidation Backstop Provider (LBP) junior capital layer
purpose: Absorbs recognized shortfall ahead of senior principal in exchange for enhanced protocol income
A new protocol role; the only support is the author's simulator and fixtures. No live deployment, no external validation.
Loss-participating withdrawal queue
purpose: Prevents first-exit advantage by keeping queued shares exposed to NAV changes until payment
Mechanism verified only against author-built fixtures (H5) and the author's simulator; no real withdrawal behavior observed.
Collateralized market-maker commitment with bond slashing
purpose: Converts revocable quote depth into reserve-substitutable execution capacity
Theorem 7.2 substitution is conditional on enforceable value transfer; no venue implements or validates the construct.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Axient: On-Chain Credit and Loss Allocation for Leveraged Event Markets: A Venue-Agnostic Protocol for Traders, Credit Providers, Market Makers, and Liquidation Backstops." pith.science (2026). https://pith.science/paper/B5FSO5WX

@misc{pith2026260800647,
  author       = {Pith},
  title        = {Pith review of: Axient: On-Chain Credit and Loss Allocation for Leveraged Event Markets: A Venue-Agnostic Protocol for Traders, Credit Providers, Market Makers, and Liquidation Backstops},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B5FSO5WX}},
  note         = {Machine review of arXiv:2608.00647}
}

read the original abstract

A physically backed leveraged event position requires real credit: if collateral C receives leverage L, the protocol supplies (L-1)C and uses the combined amount to acquire recognized event exposure. This paper develops a venue-agnostic on-chain credit architecture for that capital layer and an endogenous model of its capital market. It separates traders, Senior Credit LPs, market makers, liquidators, and Liquidation Backstop Providers; formalizes pool and debt shares, utilization- and risk-sensitive interest, collateral-locked position accounts, venue capabilities, market-maker commitments, withdrawal queues, isolated pools, non-redeemable reserves, and a deterministic loss waterfall; and models endogenous provider participation, leverage demand, liquidity withdrawal, liquidator entry, reserve replenishment, runs, and common-factor contagion. Formal results establish balanced real- and integer-unit accounting, settlement-confirmed debt priority, trader residual ownership, idempotent partial settlement, non-dilutive share issuance, junior-before-Senior impairment, loss-participating withdrawal queues, utilization-equilibrium conditions, and loss-allocation and contagion bounds. The release preserves 28 exact fixtures and 31,082 deterministic checks and adds fixed-seed agent-based experiments with 70,207,488 scalar invariant evaluations and zero failures. The experiments show that layered protection reduces but does not eliminate Senior loss, that 5x leverage materially increases capital pressure, and that market-maker capacity can raise aggregate shortfall if admission expands too quickly. Results are synthetic mechanism comparisons under author-specified behavior, not forecasts of APY, defaults, venue liquidity, or production safety.

Figures

Figures reproduced from arXiv: 2608.00647 by the authors.

**Figure 2.** Registered synthetic utilization-based borrow-rate curve. Parameters are policy inputs [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗

**Figure 3.** Canonical Axient loss path. Capital reaches the senior layer only after all earlier [PITH_FULL_IMAGE:figures/full_fig_p022_3.png] view at source ↗

Figures from the paper (14 more)

**Figure 4.** Figure 4: Synthetic queue clearance after a withdrawal request equal to 40 percent of senior [PITH_FULL_IMAGE:figures/full_fig_p027_4.png]

**Figure 5.** Figure 5: Higher leverage increases the financed share of a position and decreases gross exposure [PITH_FULL_IMAGE:figures/full_fig_p028_5.png]

**Figure 6.** Figure 6: Synthetic senior-principal loss incidence for the phased leverage mix. The values [PITH_FULL_IMAGE:figures/full_fig_p035_6.png]

**Figure 7.** Figure 7: Synthetic provider-return distributions under reserve, LBP, and bonded-MM protection. [PITH_FULL_IMAGE:figures/full_fig_p035_7.png]

**Figure 8.** Figure 8: Synthetic leverage-mix comparison. The bars use different units and should be read as [PITH_FULL_IMAGE:figures/full_fig_p036_8.png]

**Figure 9.** Figure 9: Synthetic total-network senior-loss incidence under isolated and shared protection. [PITH_FULL_IMAGE:figures/full_fig_p037_9.png]

**Figure 10.** Figure 10: Protection-demand frontier under the phased leverage policy. Both axes are synthetic [PITH_FULL_IMAGE:figures/full_fig_p046_10.png]

**Figure 11.** Figure 11: Leverage-policy trade-off under the full configuration. [PITH_FULL_IMAGE:figures/full_fig_p047_11.png]

**Figure 12.** Figure 12: Mean raw shortfall during registered strategic-MM-withdrawal weeks. [PITH_FULL_IMAGE:figures/full_fig_p047_12.png]

**Figure 13.** Figure 13: Run-week incidence and mean queue-delay proxy across configurations under phased [PITH_FULL_IMAGE:figures/full_fig_p048_13.png]

**Figure 14.** Figure 14: Six hypotheses pass and H4 fails. Values use different metric scales; the figure is a [PITH_FULL_IMAGE:figures/full_fig_p050_14.png]

**Figure 15.** Figure 15: Material Senior-loss incidence and withdrawal-run incidence in the independent daily [PITH_FULL_IMAGE:figures/full_fig_p086_15.png]

**Figure 16.** Figure 16: Synthetic withdrawal-run phase map. A cash buffer is a liquidity policy, not a [PITH_FULL_IMAGE:figures/full_fig_p086_16.png]

**Figure 17.** Figure 17: Independent daily leverage-mix trade-off. [PITH_FULL_IMAGE:figures/full_fig_p087_17.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

18 extracted references · 2 canonical work pages

[11]

Charles-Albert Lehalle and Othmane Mounjid

doi: 10.2307/1913210. Charles-Albert Lehalle and Othmane Mounjid. Limit order strategic placement with adverse selection risk and the role of latency,

doi:10.2307/1913210
[17]

Leigh Tesfatsion

arXiv:2602.12104. Leigh Tesfatsion. Agent-based computational economics: A constructive approach to economic theory. InHandbook of Computational Economics, volume 2, pages 831–880. Elsevier,

arXiv
[18]

doi: 10.1016/S1574-0021(05)02016-2. Sam M. Werner, Daniel Perez, Lewis Gudgeon, Ariah Klages-Mundt, Dominik Harz, and William J. Knottenbelt. Sok: Decentralized finance (defi). InProceedings of the 4th ACM Conference on Advances in Financial Technologies,

doi:10.1016/s1574-0021(05)02016-2
[456]

Thinking fast and slow: Data-driven adaptive defi borrow-lending protocol, 2024a

Mahsa Bastankhah, Viraj Nadkarni, Xuechao Wang, Chi Jin, Sanjeev Kulkarni, and Pramod Viswanath. Thinking fast and slow: Data-driven adaptive defi borrow-lending protocol, 2024a. arXiv:2407.10890. Mahsa Bastankhah, Viraj Nadkarni, Xuechao Wang, and Pramod Viswanath. Agilerate: Bringing adaptivity and robustness to defi lending markets, 2024b. arXiv:2410.1...

arXiv
[1985]

Itay Goldstein and Ady Pauzner

doi: 10.1016/0304-405X(85)90044-3. Itay Goldstein and Ady Pauzner. Demand-deposit contracts and the probability of bank runs. Journal of Finance, 60(3):1293–1327,

doi:10.1016/0304-405x(85)90044-3
[1999]

Charles F

doi: 10.1038/17290. Charles F. Manski. Interpreting the predictions of prediction markets.Economics Letters, 91(3): 425–429,

doi:10.1038/17290
[2005]

Lewis Gudgeon, Daniel Perez, Dominik Harz, Benjamin Livshits, and Arthur Gervais

doi: 10.1111/j.1540-6261.2005.00762.x. Lewis Gudgeon, Daniel Perez, Dominik Harz, Benjamin Livshits, and Arthur Gervais. The decentralized financial crisis. In2020 Crypto Valley Conference on Blockchain Technology,

arXiv 2005
[2006]

Axient: Debt-free finality for leveraged binary event markets, 2026a

Maksym Nechepurenko. Axient: Debt-free finality for leveraged binary event markets, 2026a. Axient research manuscript, version r0.3.1. Maksym Nechepurenko. Fill-side non-retail trading on Polymarket: An empirical study of behavioral tiers and microstructure signatures under quote-attribution constraints, 2026b. arXiv:2605.11640. Maksym Nechepurenko. Manip...

arXiv

Show all 18 references

[2007]

doi: 10.1007/s10614-007-9106-0. J. Doyne Farmer and Duncan Foley. The economy needs agent-based modelling.Nature, 460: 685–686,

doi
[2009]

88 Pietro Fodra and Mauricio Labadie

doi: 10.1038/460685a. 88 Pietro Fodra and Mauricio Labadie. High-frequency market-making with inventory constraints and directional bets,

doi
[2010]

Lawrence R

doi: 10.1098/rspa.2009.0410. Lawrence R. Glosten and Paul R. Milgrom. Bid, ask and transaction prices in a specialist market with heterogeneously informed traders.Journal of Financial Economics, 14(1):71–100,

2009
[2012]

Prasanna Gai and Sujit Kapadia

arXiv:1206.4810. Prasanna Gai and Sujit Kapadia. Contagion in financial networks.Proceedings of the Royal Society A, 466(2120):2401–2423,

arXiv
[2015]

doi: 10.1257/aer.2013

2013 doi
[2016]

Robert Leshner and Geoffrey Hayes

arXiv:1610.00261. Robert Leshner and Geoffrey Hayes. Compound: The money market protocol. White paper,

arXiv
[2020]

Kaihua Qin, Jens Ernstberger, Liyi Zhou, Philipp Jovanovic, and Arthur Gervais

arXiv:2009.13235. Kaihua Qin, Jens Ernstberger, Liyi Zhou, Philipp Jovanovic, and Arthur Gervais. Mitigating decentralized finance liquidations with reversible call options,

2009 arXiv
[2022]

Douglas W

arXiv:2211.08870. Douglas W. Diamond and Philip H. Dybvig. Bank runs, deposit insurance, and liquidity.Journal of Political Economy, 91(3):401–419,

arXiv
[2023]

89 Agathe Sadeghi and Zachary Feinstein

arXiv:2303.15162. 89 Agathe Sadeghi and Zachary Feinstein. Liquidation dynamics in defi and the role of transaction fees,

arXiv
[2026]

Daron Acemoglu, Asuman Ozdaglar, and Alireza Tahbaz-Salehi

Accessed 2026-07-14; background for pooled lending, utilization rates, reserve factors, and liquidation design. Daron Acemoglu, Asuman Ozdaglar, and Alireza Tahbaz-Salehi. Systemic risk and stability in financial networks.American Economic Review, 105(2):564–608,

2026

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.

[1] [11]

Charles-Albert Lehalle and Othmane Mounjid

doi: 10.2307/1913210. Charles-Albert Lehalle and Othmane Mounjid. Limit order strategic placement with adverse selection risk and the role of latency,

doi:10.2307/1913210

[2] [17]

Leigh Tesfatsion

arXiv:2602.12104. Leigh Tesfatsion. Agent-based computational economics: A constructive approach to economic theory. InHandbook of Computational Economics, volume 2, pages 831–880. Elsevier,

arXiv

[3] [18]

doi: 10.1016/S1574-0021(05)02016-2. Sam M. Werner, Daniel Perez, Lewis Gudgeon, Ariah Klages-Mundt, Dominik Harz, and William J. Knottenbelt. Sok: Decentralized finance (defi). InProceedings of the 4th ACM Conference on Advances in Financial Technologies,

doi:10.1016/s1574-0021(05)02016-2

[4] [456]

Thinking fast and slow: Data-driven adaptive defi borrow-lending protocol, 2024a

Mahsa Bastankhah, Viraj Nadkarni, Xuechao Wang, Chi Jin, Sanjeev Kulkarni, and Pramod Viswanath. Thinking fast and slow: Data-driven adaptive defi borrow-lending protocol, 2024a. arXiv:2407.10890. Mahsa Bastankhah, Viraj Nadkarni, Xuechao Wang, and Pramod Viswanath. Agilerate: Bringing adaptivity and robustness to defi lending markets, 2024b. arXiv:2410.1...

arXiv

[5] [1985]

Itay Goldstein and Ady Pauzner

doi: 10.1016/0304-405X(85)90044-3. Itay Goldstein and Ady Pauzner. Demand-deposit contracts and the probability of bank runs. Journal of Finance, 60(3):1293–1327,

doi:10.1016/0304-405x(85)90044-3

[6] [1999]

Charles F

doi: 10.1038/17290. Charles F. Manski. Interpreting the predictions of prediction markets.Economics Letters, 91(3): 425–429,

doi:10.1038/17290

[7] [2005]

Lewis Gudgeon, Daniel Perez, Dominik Harz, Benjamin Livshits, and Arthur Gervais

doi: 10.1111/j.1540-6261.2005.00762.x. Lewis Gudgeon, Daniel Perez, Dominik Harz, Benjamin Livshits, and Arthur Gervais. The decentralized financial crisis. In2020 Crypto Valley Conference on Blockchain Technology,

arXiv 2005

[8] [2006]

Axient: Debt-free finality for leveraged binary event markets, 2026a

Maksym Nechepurenko. Axient: Debt-free finality for leveraged binary event markets, 2026a. Axient research manuscript, version r0.3.1. Maksym Nechepurenko. Fill-side non-retail trading on Polymarket: An empirical study of behavioral tiers and microstructure signatures under quote-attribution constraints, 2026b. arXiv:2605.11640. Maksym Nechepurenko. Manip...

arXiv