Corrective Retrieval Augmented Generation (CRAG) improves the robustness of RAG systems by evaluating retrieved document quality and triggering corrective actions. However, the original implementation relies on proprietary components including the Google Search API and closed model weights, limiting reproducibility. In this work, we present a fully open-source reproduction of CRAG, replacing proprietary web search with the Wikipedia API and the original LLaMA-2 generator with Phi-3-mini-4k-instruct. We evaluate on PopQA and ARC-Challenge, demonstrating that our open-source pipeline achieves comparable performance to the original system. Furthermore, we contribute the first explainability analysis of CRAG's T5-based retrieval evaluator using SHAP, revealing that the evaluator primarily relies on named entity alignment rather than semantic similarity. Our analysis identifies key failure modes including domain transfer limitations on science questions. All code and results are available at https://github.com/suryayalavarthi/crag-reproduction.
In the development and maintenance of Android apps, the quick and accurate reproduction of user-reported bugs is crucial to ensure application quality and improve user satisfaction. However, this process is often time-consuming and complex. Therefore, there is a need for an automated approach that can explore the Application Under Test (AUT) and identify the correct sequence of User Interface (UI) actions required to reproduce a bug, given only a complete bug report. Large Language Models (LLMs) have shown remarkable capabilities in understanding textual and visual semantics, making them a promising tool for planning UI actions. Nevertheless, our study shows that even when using state-of-the-art LLM-based approaches, these methods still struggle to follow detailed bug reproduction instructions and replan based on new information, due to their inability to accurately predict and interpret the visual effects of UI components. To address these limitations, we propose LTGDroid. Our insight is to execute all possible UI actions on the current UI page during exploration, record their corresponding visual effects, and leverage these visual cues to guide the LLM in selecting UI actions that are likely to reproduce the bug. We evaluated LTGDroid, instantiated with GPT-4.1, on a benchmark consisting of 75 bug reports from 45 popular Android apps. The results show that LTGDroid achieves a reproduction success rate of 87.51%, improving over the state-of-the-art baselines by 49.16% and 556.30%, while requiring an average of 20.45 minutes and approximately $0.27 to successfully reproduce a bug. The LTGDroid implementation is publicly available at https://github.com/N3onFlux/LTGDroid.
ABSTRACTOdourant receptors (ORs) are not restricted only to the nose, but also occur in many other organs and tissues, including the reproductive system. In fact, ORs are the most heavily expressed in testis than in any other extra‐nasal tissue. Accumulating evidence suggests that olfactory and reproductive systems are both structurally and functionally linked and that these interconnections can influence various aspects of reproduction. In this article, we first review our current understanding of these interconnections and then collate accumulated evidence on the presence of ORs in the male reproductive system and sperm cells. We then investigate the potential role of female reproductive tract odourants in sperm chemotaxis and selection. Finally, since the existing evidence especially for sperm odor sensing capability and its physiological function are controversial, we also review potential reasons for the controversy and propose some ways to resolve the debate. Collectively, we conclude that reproductive odourant signaling may play an important, although currently largely unclear role in many key processes directly related to male fertility. However, since we lack holistic understanding of the functional significance of ORs and odor sensing pathways of the male reproductive system, more empirical research is warranted.
In the face of an infectious disease, a key epidemiological measure is the basic reproduction number, which quantifies the average secondary infections caused by a single case in a susceptible population. In practice, the effective reproduction number, denoted as $R_t$, is widely used to assess the transmissibility of the disease at a given time $t$. Real-time estimating this metric is vital for understanding and managing disease outbreaks. Traditional statistical inference often relies on two assumptions. One is that samples are assumed to be drawn from a homogeneous population distribution, neglecting significant variations in individual transmission rates. The other is the ideal case reporting assumption, disregarding time delays between infection and reporting. In this paper, we thoroughly investigate these critical factors and assess their impact on estimating $R_t$. We first introduce negative binomial and Weibull distributions to characterize transmission rates and reporting delays, respectively, based on which observation and state equations are formulated. Then, we employ a Bayesian filtering for estimating $R_t$. Finally, validation using synthetic and empirical data demonstrates a significant improvement in estimation accuracy compared to conventional methods that ignore these factors.
We introduce the boundary reproduction number, adapted from the next generation matrix method, to assess whether an infusion of species will persist or become exhausted in a chemical reaction system. Our main contributions are as follows: (a) we show how the concept of a siphon, prevalent in Petri nets and chemical reaction network theory, identifies sets of species that may become depleted at steady state, analogous to a disease-free boundary steady state; (b) we develop an approach for incorporating biochemically motivated conservation laws, which allows the stability of boundary steady states to be determined within specific compatibility classes; and (c) we present an effective heuristic for decomposing the Jacobian of the system that reduces the computational complexity required to compute the stability domain of a boundary steady state. The boundary reproduction number approach significantly simplifies existing parameter-dependent methods for determining the stability of boundary steady states in chemical reaction systems and has implications for the capacity of critical metabolites and substrates in metabolic pathways to become exhausted.
Mathematical expressions play a central role in scientific discovery. Symbolic regression aims to automatically discover such expressions from given numerical data. Recently, Neural symbolic regression (NSR) methods that involve Transformers pre-trained on synthetic datasets have gained attention for their fast inference, but they often perform poorly, especially with many input variables. In this study, we analyze NSR from both theoretical and empirical perspectives and show that (1) ordinary token-by-token generation is ill-suited for NSR, as Transformers cannot compositionally generate tokens while validating numerical consistency, and (2) the search space of NSR methods is greatly restricted due to reproduction bias, where the majority of generated expressions are merely copied from the training data. We further examine whether tailored test-time strategies can reduce reproduction bias and show that providing additional information at test time effectively mitigates it. These findings contribute to a deeper understanding of the limitation of NSR approaches and provide guidance for designing more robust and generalizable methods. Code is available at https://github.com/Shun-0922/Mem-Bias-NSR .
Basic and instantaneous reproduction numbers, "R" _"0" and "R" _"t" , are important metrics to assess progress of an epidemic and effectiveness of preventative interventions undertaken, and also to estimate coverage needed for vaccination. Reproduction numbers are related to the daily number of positive cases recorded by the national public health authorities, via the renewal equation. During periods of exponential growth or decay they are linked also to the rate constants by the Lotka-Euler equation. For either application, we need the distribution of generation times between primary and secondary infections. In practice, we use instead the directly observable serial interval between symptoms onset of infector and infectee. Pre-symptomatic transmission that occurs in COVID infection causes serial intervals to extend to negative values, which can be described with a Gaussian distribution. Consistent application of the two approaches requires careful attention to lower limits imposed on the distribution. Allowing Gaussian-distributed serial intervals to extend to minus infinity with the Lotka-Euler equation, as commonly is done, results in lower reproduction numbers than predicted from the discretized renewal equation. Here, we formulate the Lotka-Euler equation for Gaussian distributions including an explicit lower cut-off, and use this to explore the consequences of presymptomatic transmission for COVID-19 infections.
A class of multiple-timescale asymptotic solutions to the equations of the susceptible-infected-recovered (SIR) model is presented for the case of high basic reproduction number, with the inverse of the latter employed as the expansion parameter. High values of the basic reproduction number, a coefficient defined as the ratio of the infection and recovery rates built into the SIR model equations, are associated with escalating epidemics. Combinations of fast and slow timescales in the suggested multiple-timescale solutions prove adequate to reflect the acknowledged epidemic paradigm, which is characterized by the concatenation of a sharp outbreak with a subsequent protracted plateau. Explicit solutions for the numbers of the infected, susceptible, and recovered compartments of the SIR model are derived via the asymptotic treatment, and the epidemic peak timing and magnitude are assessed on this basis. The asymptotic results agree seamlessly with numerical simulations based on the SIR model.
Sexual interactions have previously been shown to improve reproductive health through unknown mechanisms. In this study, we used RNA-Seq to examine sex-induced gene expression changes in the preoptic area (POA), a critical reproductive brain region. Using a mouse model defective in fibroblast growth factor signaling (dnFGFR mouse), previously shown to disrupt the gonadotropin-releasing hormone (GnRH) system, we examined the impact of opposite sex (OS) housing on gene expression in the POA of a reproductively compromised animal. Bulk RNA-Seq followed by gene set enrichment analysis (GSEA) were used to analyze changes in gene expression and biological processes in control and dnFGFR mice after 300 days of cohabitation with a same sex or OS partner. OS housing of dnFGFR mice, but not control mice, significantly improved reproductive anatomy and gonadotropins in dnFGFR mice. These changes occurred concomitantly with novel biological processes related to estradiol metabolism and neuron excitation. Our results suggest a new role of neuron- or astrocyte-derived estradiol in the plasticity of the GnRH neuron population and offer a promising new direction for the treatment of reproductive disorders stemming from GnRH deficiency.
Diseases of the endocrine glands. Clinical endocrinology