Selected work

Representative Papers

A Machine-checked Proof of Consistency for Impredicative Pure Type Systems

Jul 22, 2026

This work addresses the consistency problem of pure type systems (PTS) featuring impredicativity by proposing a formal method grounded in α-conversion relations and Stoughton’s notion of parallel substitution. Building upon Takahashi’s refinement of the Tait–Martin-Löf normalization technique, the authors fully mechanize in Agda the confluence of β-reduction and subject reduction properties, and—under a normalization assumption—achieve the first machine-checked proof of logical consistency for a subclass of impredicative PTS. The study not only clarifies subtle technical challenges in establishing confluence but also demonstrates both the feasibility and inherent limitations of mechanized consistency proofs for higher-order type theories.

0 citationsRead paper

\chisao{}: A GPU-Native Parallel Optimizer for Multimodal Black-Box Functions via Convergence-Anticonvergence Oscillation

Jun 24, 2026

This work addresses the challenge of efficiently locating global extrema of multimodal black-box functions by proposing a population-based optimizer natively designed for GPU parallelization. The method introduces an asymmetric oscillation mechanism that dynamically balances convergence and anti-convergence, freezing discovered optima while driving remaining samples to persistently explore the search space. It integrates momentum-based anti-convergence, stochastic smoothed gradient estimation, and adaptive resampling strategies—dubbed Repulse Monkey and Golden Rooster—operating entirely without gradient information. Evaluated on all 42 high-dimensional (up to 64D) multimodal functions in the SFU benchmark suite, the algorithm achieves 100% basin recovery, attains up to 39× speedup over basin-hopping, and demonstrates robust performance under strong noise, significantly outperforming existing CPU-based baselines.

0 citationsRead paper

Minimality of Random Moore Automata under Prefix-Dependent Congruences

Jun 18, 2026

This study addresses the minimization of stochastic deterministic transition systems under prefix-dependent congruence, where the set of admissible future inputs depends on the observed input prefix, and two states are considered equivalent if no legal input sequence can distinguish them. Focusing on stochastic deterministic automata with state outputs—whose transitions and labels are drawn independently from uniform distributions—the work establishes, for the first time in this setting, that when the label consistency probability is strictly less than one and each state admits at least three admissible symbols, the induced congruence relation is trivial with high probability (i.e., all states are pairwise inequivalent). By integrating pairwise pruning, collision-free exploration to control early evolution, and a first-moment analysis of surviving state pairs, the authors demonstrate that under these conditions the system is almost surely non-minimizable.

0 citationsRead paper

The Markup falsification Adaptative Set

May 28, 2026

This study addresses the failure of the markup estimation method proposed by De Loecker and Warzynski (2012) when its key identifying assumptions are falsified. To overcome this limitation, the paper introduces a general framework that requires no additional assumptions: by continuously relaxing the standard identifying conditions, it constructs a non-empty set of non-falsified models and expresses markups as a function of the relaxation parameters, thereby generating an adaptive identification set. This approach extends conventional point identification to interval identification that reflects potential violations of underlying assumptions. Applying this method to Chilean firm-level data from Raval (2023), the empirical analysis successfully constructs robust markup identification sets, effectively mitigating the identification risks inherent in the original approach. The framework thus offers a novel contribution by enhancing model robustness while preserving identification rigor.

0 citationsRead paper

Stochastic Frontier meets Breakdown Frontier

Apr 28, 2026

This study addresses the sensitivity of average production inefficiency estimates in stochastic frontier models to benchmark assumptions by introducing, for the first time, the breakdown frontier approach into this framework. By relaxing key identifying assumptions, the paper characterizes the identified set of parameters and derives the breakdown frontier for the parameter of interest to quantify the boundary of misspecification bias. Integrating identified set analysis with sensitivity analysis techniques, the proposed method is validated on classical empirical datasets, demonstrating its effectiveness in assessing robustness. To facilitate reproducibility and practical application, the authors publicly release their implementation code, providing researchers and practitioners with a transparent and accessible tool for evaluating the robustness of inefficiency estimates under potential model misspecification.

0 citationsRead paper

Recent publications

Latest Papers

A Machine-checked Proof of Consistency for Impredicative Pure Type Systems

Jul 22, 2026

This work addresses the consistency problem of pure type systems (PTS) featuring impredicativity by proposing a formal method grounded in α-conversion relations and Stoughton’s notion of parallel substitution. Building upon Takahashi’s refinement of the Tait–Martin-Löf normalization technique, the authors fully mechanize in Agda the confluence of β-reduction and subject reduction properties, and—under a normalization assumption—achieve the first machine-checked proof of logical consistency for a subclass of impredicative PTS. The study not only clarifies subtle technical challenges in establishing confluence but also demonstrates both the feasibility and inherent limitations of mechanized consistency proofs for higher-order type theories.

0 citationsRead paper

\chisao{}: A GPU-Native Parallel Optimizer for Multimodal Black-Box Functions via Convergence-Anticonvergence Oscillation

Jun 24, 2026

This work addresses the challenge of efficiently locating global extrema of multimodal black-box functions by proposing a population-based optimizer natively designed for GPU parallelization. The method introduces an asymmetric oscillation mechanism that dynamically balances convergence and anti-convergence, freezing discovered optima while driving remaining samples to persistently explore the search space. It integrates momentum-based anti-convergence, stochastic smoothed gradient estimation, and adaptive resampling strategies—dubbed Repulse Monkey and Golden Rooster—operating entirely without gradient information. Evaluated on all 42 high-dimensional (up to 64D) multimodal functions in the SFU benchmark suite, the algorithm achieves 100% basin recovery, attains up to 39× speedup over basin-hopping, and demonstrates robust performance under strong noise, significantly outperforming existing CPU-based baselines.

0 citationsRead paper

Minimality of Random Moore Automata under Prefix-Dependent Congruences

Jun 18, 2026

This study addresses the minimization of stochastic deterministic transition systems under prefix-dependent congruence, where the set of admissible future inputs depends on the observed input prefix, and two states are considered equivalent if no legal input sequence can distinguish them. Focusing on stochastic deterministic automata with state outputs—whose transitions and labels are drawn independently from uniform distributions—the work establishes, for the first time in this setting, that when the label consistency probability is strictly less than one and each state admits at least three admissible symbols, the induced congruence relation is trivial with high probability (i.e., all states are pairwise inequivalent). By integrating pairwise pruning, collision-free exploration to control early evolution, and a first-moment analysis of surviving state pairs, the authors demonstrate that under these conditions the system is almost surely non-minimizable.

0 citationsRead paper

The Markup falsification Adaptative Set

May 28, 2026

This study addresses the failure of the markup estimation method proposed by De Loecker and Warzynski (2012) when its key identifying assumptions are falsified. To overcome this limitation, the paper introduces a general framework that requires no additional assumptions: by continuously relaxing the standard identifying conditions, it constructs a non-empty set of non-falsified models and expresses markups as a function of the relaxation parameters, thereby generating an adaptive identification set. This approach extends conventional point identification to interval identification that reflects potential violations of underlying assumptions. Applying this method to Chilean firm-level data from Raval (2023), the empirical analysis successfully constructs robust markup identification sets, effectively mitigating the identification risks inherent in the original approach. The framework thus offers a novel contribution by enhancing model robustness while preserving identification rigor.

0 citationsRead paper

Stochastic Frontier meets Breakdown Frontier

Apr 28, 2026

This study addresses the sensitivity of average production inefficiency estimates in stochastic frontier models to benchmark assumptions by introducing, for the first time, the breakdown frontier approach into this framework. By relaxing key identifying assumptions, the paper characterizes the identified set of parameters and derives the breakdown frontier for the parameter of interest to quantify the boundary of misspecification bias. Integrating identified set analysis with sensitivity analysis techniques, the proposed method is validated on classical empirical datasets, demonstrating its effectiveness in assessing robustness. To facilitate reproducibility and practical application, the authors publicly release their implementation code, providing researchers and practitioners with a transparent and accessible tool for evaluating the robustness of inefficiency estimates under potential model misspecification.

0 citationsRead paper