Search for: All records

High‑quality Question–Answer (QA) datasets are foundational for reliable Large Language Model (LLM) evaluation, yet even expert‑crafted datasets exhibit persistent gaps in domain coverage, misaligned difficulty distributions, and factual inconsistencies. The recent surge in generative model-powered datasets has compounded these quality challenges. In this work, we introduce RefineLab, the first LLM‑driven framework that automatically refines raw QA textual data into high-quality datasets under a controllable token‑budget constraint. RefineLab takes a set of target quality attributes as refinement objectives and performs selective edits within a predefined token budget to ensure practicality and efficiency. In essence, RefineLab addresses a constrained optimization problem: improving the quality of QA samples as much as possible while respecting resource limitations. With a set of available refinement operations, RefineLab takes as input the original dataset, a specified set of target quality dimensions, and a token budget, and determines which refinement operations should be applied to each QA sample. This process is guided by an assignment module that selects optimal refinement strategies to maximize overall dataset quality while adhering to the budget constraint. Experiments demonstrate that RefineLab consistently narrows divergence from expert datasets across coverage, difficulty alignment, factual fidelity, and distractor quality. RefineLab pioneers a scalable, customizable path to reproducible dataset design, with broad implications for LLM evaluation. 

Abstract Recent interest in urban and regional air mobility and the need to improve the aviation industry’s emissions has motivated research and development of novel propeller-driven vehicles. These vehicles range in configuration from conventional takeoff and landing designs to complex rotorcraft that transition between vertical and horizontal flight. These designs must be optimized to ensure optimal efficiency throughout their missions, leveraging the tightly coupled nature of propeller-wing interaction. In this work, we study the NASA tiltwing concept vehicle wing with varying numbers of propellers, ranging from no propellers to five propellers evenly spaced along the wing. Using aerodynamic shape optimization, we optimize the wing shapes for each propeller-wing configuration, minimizing the wing drag. These optimizations are carried out with DAFoam, a discrete adjoint implementation of OpenFOAM, embedded within OpenMDAO and the MPhys optimization framework. The optimizations show that the lowest drag configuration is a single propeller mounted at the wing tip. Increasing the number of propellers slightly increases drag compared to the single propeller configuration. However, aerodynamic shape optimization considering propeller-wing interaction yields a negligible benefit compared to aerodynamic optimization of an isolated wing that is subsequently trimmed to a desired flight condition in the presence of a propeller. 

Abstract BackgroundAlternative splicing of precursor mRNAs serves as a crucial mechanism to enhance gene expression plasticity for organismal adaptation. However, the precise regulation and function of alternative splicing in plant immune gene regulation remain elusive. ResultsHere, by deploying in-depth transcriptome profiling with deep genome coverage coupled with differential expression, differential alternative splicing, and differential transcript usage analysis, we reveal profound and dynamic changes in alternative splicing following treatment with microbial pattern flg22 peptides inArabidopsis. Our findings highlight RNA polymerase II C-terminal domain phosphatase-like 3 (CPL3) as a key regulator of alternative splicing, preferentially influencing the splicing patterns of defense genes rather than their expression levels. CPL3 mediates the production of a flg22-induced alternative splicing variant, diacylglycerol kinase 5α (DGK5α), which differs from the canonical DGK5β in its interaction with the upstream kinase BIK1 and subsequent phosphorylation, resulting in reduced flg22-triggered production of phosphatidic acid and reactive oxygen species. Furthermore, our functional analysis suggests that DGK5β, but not DGK5α, contributes to plant resistance against virulent and avirulent bacterial infections. ConclusionsThese findings underscore the role of CPL3 in modulating alternative splicing dynamics of defense genes and DGK5 isoform-mediated phosphatidic acid homeostasis, shedding light on the intricate mechanisms underlying plant immune gene regulation. 

Abstract In plant immunity, a well-orchestrated cascade is initiated by the dimerization of receptor-like kinases (RLKs), followed by the phosphorylation of receptor-like cytoplasmic kinases (RLCKs) and subsequent activation of NADPH oxidases for ROS generation. Recent findings by Zhong et al. illustrated that a maize signaling module comprising ZmWAKL-ZmWIK-ZmBLK1-ZmRBOH4 governs quantitative disease resistance to grey leaf spot, a pervasive fungal disease in maize worldwide, unveiling the conservation of this signaling quartet in plant immunity.