Jinjie Liu

Jinyao Guo, Chengpeng Wang, Dominic Deluca et al.

Software auditing is an increasingly critical task in the era of rapid code generation. While LLM-based auditors have demonstrated strong potential, their effectiveness remains limited by misalignment with the highly complex, domain-specific nature of bug detection. In this work, we introduce BugScope, a framework that mirrors how human auditors learn specific bug patterns from representative examples and apply this knowledge during code auditing. BugScope structures auditing into three steps: seed identification, context retrieval, and bug detection, and aligns LLMs to each step by analyzing real bug reports and mutated examples, and distilling concise, reusable guidelines. On a curated dataset of 33 real-world bugs from 21 widely used open-source projects, BugScope achieves 86.05\% precision and 87.88\% recall, corresponding to an F1 score of 0.87. By comparison, leading industrial tools such as Claude Code (with Claude Opus 4.6) and Cursor BugBot achieve F1 scores of only 0.51 and 0.43, respectively. Beyond benchmarks, large-scale evaluation on real-world projects such as the Linux kernel uncovered 184 previously unknown bugs, of which 78 have already been fixed and 7 explicitly confirmed by developers. Our code is available at https://github.com/jinyaoguo/BugScope

Jinjie Liu, Moysey Brio

This paper investigates the energy conservation properties of explicit Runge--Kutta (RK) time discretizations for autonomous skew-symmetric systems. For linear problems, we present a general framework for constructing RK methods in which the energy-accuracy order significantly exceeds the number of stages. Specifically, for an $s$-stage, $p$-th order RK method (where $p$ is even), we prove that the energy accuracy can reach up to order $2s-p+1$. Utilizing this framework, we derive several energy-superconvergent methods, including five- to seven-stage algorithms with energy accuracy up to the eleventh order, and establish their corresponding strong stability criteria. The methods are validated on a range of benchmark problems, including harmonic oscillators, integro-differential equations in peridynamics, and the Maxwell equations. Furthermore, we extend the energy-superconvergent framework to autonomous nonlinear systems with amplitude-dependent frequencies. By deriving fifth-order energy conditions for three-stage, second-order methods, we develop the RK325 algorithm. The performance of RK325 is demonstrated for a broad range of problems, including Euler's equations for rigid body dynamics, the nonlinear Schrödinger equation, the Korteweg--de Vries (KdV) equation, Burgers' equation, and the Landau--Lifshitz equation. Additionally, we develop four-stage, second-order methods (RK427) and five-stage, fourth-order methods (RK547), all of which achieve seventh-order energy accuracy for the cubic nonlinear case. Finally, the performance of RK547 method is illustrated using the nonlinear Maxwell--Kerr system.

Zhu Deng, Jinjie Liu, Biao Luo et al.

The product carbon footprint (PCF) is crucial for decarbonizing the supply chain, as it measures the direct and indirect greenhouse gas emissions caused by all activities during the product's life cycle. However, PCF accounting often requires expert knowledge and significant time to construct life cycle models. In this study, we test and compare the emergent ability of five large language models (LLMs) in modeling the 'cradle-to-gate' life cycles of products and generating the inventory data of inputs and outputs, revealing their limitations as a generalized PCF knowledge database. By utilizing LLMs, we propose an automatic AI-driven PCF accounting framework, called AutoPCF, which also applies deep learning algorithms to automatically match calculation parameters, and ultimately calculate the PCF. The results of estimating the carbon footprint for three case products using the AutoPCF framework demonstrate its potential in achieving automatic modeling and estimation of PCF with a large reduction in modeling time from days to minutes.

Jinjie Liu, Hang Qiu

Machine learning models deployed on edge devices have enabled numerous exciting new applications, such as humanoid robots, AR glasses, and autonomous vehicles. However, the computing resources available on these edge devices are not catching up with the ever-growing number of parameters in these models. As the models become bigger and more complicated, the novel yet sophisticated structure challenges the inference runtime optimization. We present FluidML, a generic runtime memory management and optimization framework that can flexibly transform the model execution blueprint to achieve faster and more memory-efficient inference. Evaluations across different platforms show that FluidML can consistently reduce the end-to-end inference latency by up to 25.38% for popular language models and reduce peak memory usage by up to 41.47%, compared to state-of-the-art approaches. FluidML is of ~30K line of codes, built for general-purpose usage, and will be released as an open-source inference runtime optimization framework to the community.

Qiqi Tran, Jinjie Liu

The iterated Crank-Nicolson (ICN) method is a successful numerical algorithm in numerical relativity for solving partial differential equations. The $θ$-ICN method is the extension of the original ICN method where $θ$ is the weight when averaging the predicted and corrected values. It has better stability when $θ$ is chosen to be larger than 0.5, but the accuracy is reduced since the $θ$-ICN method is second order accurate only when $θ$ = 0.5. In this paper, we propose two modified $θ$-ICN algorithms that have second order of convergence rate when $θ$ is not 0.5, based on two different ways to choose the weight $θ$. The first approach employs two geometrically averaged $θ$s in two iterations within one time step, and the second one uses arithmetically averaged $θ$s for two consecutive time steps while $θ$ remains the same in each time step. The stability and second order accuracy of our methods are verified using stability and truncation error analysis and are demonstrated by numerical examples on linear and semi-linear hyperbolic partial differential equations and Burgers' equation.