arXiv subjects

Rafael Gomes

Publications and source records attributed to Rafael Gomes.

5 recordsLinked to original sources

arXiv2026

Integrating Crash Report Mining and LLMs for Bug Localization and Repair: An Industrial Report

Analyzing crash-report bugs in large-scale industrial software systems requires substantial maintenance effort, particularly in production environments where developers must handle large volumes of crash reports and source code artifacts to localize and fix their root causes. While recent studies have shown that Large Language Models (LLMs) can assist with maintenance tasks, little is known about their effectiveness in supporting developers in analyzing crash-report bugs and repairing bugs associated with groups of crash reports in industrial settings. To address this gap, we investigate whether integrating crash report mining techniques---specifically stack trace clustering and suspicious file and method ranking---with LLMs can support crash localization and repair in production environments. We conduct a retrospective evaluation of five LLMs under four prompt configurations. After that, we chose the best model to run on 38 crash bugs collected from two large Java enterprise systems. We further analyze the structural characteristics and explanatory patterns of LLM-generated responses and assess localization and repair effectiveness through manual validation. Our results show that the best-performing configuration localizes up to 71% and correctly repairs 52% of crash bugs on the full dataset. These findings provide empirical evidence that combining crash report mining with LLM-based repair can effectively support debugging activities in industrial maintenance workflows.

cs.SE↗

arXiv2026

Improved Lower Bounds and Output Augmentation for Facility Location Mechanisms

We study the strategic facility location problem under the egalitarian objective, where a mechanism uses the reported locations of a set of agents in Euclidean space to select a facility location that minimizes the maximum distance to any agent. We restrict our attention to strategyproof mechanisms, ensuring that no agent can benefit from misreporting their location. As our main results, we prove an asymptotic lower bound of $1 + \sqrt{d/(2(d+1))}$ on the approximation ratio of any mechanism that is strategyproof in expectation in $\mathbb{R}^d$. We show that this barrier is driven by large populations by providing a randomized $\sqrt{2}$-approximate mechanism for the two-agent case. We then consider an output-augmented framework, which allows the facility to be placed outside the agents' restricted domain. For the setting where agents are restricted to a line but the facility can be anywhere in the plane, we design a deterministic strategyproof $\sqrt{2}$-approximate mechanism with a matching lower bound, showing that output augmentation can replace the need for randomness. For the setting where the agents' reports lie on the unit circle but the facility can be placed anywhere in $\mathbb{R}^2$ we introduce a randomized $3/2$-approximate mechanism that is group-strategyproof in expectation.

cs.GT↗

arXiv2025

Rigidifying simplicial complexes and realizing group actions

We show that any action of a finite group on a finitely presentable group arises as the action of the group of self-homotopy equivalences of a space on its fundamental group. In doing so, we prove that any finite connected (abstract) simplicial complex $\mathbf{K}$ can be rigidified -- meaning it can be perturbed in a way that reduces the full automorphism group to any subgroup -- while preserving the homotopy type of the geometric realization $| \mathbf{K} |$. We also obtain that every action of a finite group on a finitely generated abelian group is the action of the group of self-homotopy equivalences of a space on one of its higher homotopy groups.

math.AT↗

arXiv2024

Realization of permutation modules via Alexandroff spaces

We raise the question of the realizability of permutation modules in the context of Kahn's realizability problem for abstract groups and the $G$-Moore space problem. Specifically, given a finite group $G$, we consider a collection $\{M_i\}_{i=1}^n$ of finitely generated $\Z G$-modules that admit a submodule decomposition on which $G$ acts by permuting the summands. Then we prove the existence of connected finite spaces $X$ that realize each $M_i$ as its $i$-th homology, $G$ as its group of self-homotopy equivalences $\E(X)$, and the action of $G$ on each $M_i$ as the action of $\E(X)$ on $H_i(X; \Z)$.

math.AT↗

arXiv2021

The cohomological index of free $\Z/p$-actions is not additive with respect to join

Let $p$ denote an odd prime. We show by example that the inequalities obtained in arXiv:1704.05827 for the behaviour of the cohomological index of a join of free $\Z/p$-actions are sharp. Namely, for all odd integers $k,l$ at least one of which is greater than one, we give examples of finite free $\Z/p$-CW complexes of cohomological indices $k$ and $l$ whose join has index $k+l+1$ and also examples where the join has index $k+l-1$.

math.AT↗