arXiv subjects

Bingxin Zhao

Publications and source records attributed to Bingxin Zhao.

At least 19 recordsLinked to original sources

arXiv2026

Multi-Turn Agentic Scientific Literature Search via Workflow Induction

Scientific literature search often requires more than retrieving papers from a single query: users' intents are underspecified, preference-dependent, and evolve through interaction. Existing search agents typically rely on fixed pipelines or implicit language-only reasoning, making their search strategies difficult to control, inspect, and refine. We introduce PaperPilot, a multi-turn literature search agent that frames scientific search as workflow induction. Given an anchor paper and a user query, PaperPilot constructs an executable DAG of paper-search operators, including keyword search, citation expansion, filtering, scoring, reranking, and evidence extraction. User feedback is then used to refine both the query and the workflow itself. We train PaperPilot with supervised workflow imitation and preference optimization over controlled workflow corruptions. Experiments show that PaperPilot-9B improves over the base Qwen3.5-9B toolset agent under multi-turn interaction, increasing Hit@5 from 58.0 to 77.0, MRR from 47.5 to 59.4, and nDCG@10 from 26.8 to 32.5, while reducing workflow execution errors from 9.5% to 0%. These results show that explicit, editable search workflows provide an effective and controllable interface for aligning literature search agents with complex scientific intent.

cs.CL↗

arXiv2026

Statistical Proof as a Window into Human-AI Collaboration: Practical Insights and a Community Agenda

Large language models (LLMs) are increasingly woven into expert cognitive work in daily research, yet we know little about how human expertise should adapt when an AI system can execute substantial technical reasoning on its own. Here we use statistical proof development, a demanding and structured form of expert reasoning, as a window into this broader question. Drawing on day-to-day proof problems, we find that current general-purpose LLMs occupy a useful but limited role: they can execute technical components given a precisely formulated problem and targeted guidance, but become unreliable when the problem is open-ended or requires a long reasoning chain with multiple interdependent steps. This execution-strategy gap is rooted in what makes research-level statistical proof distinctive: unlike pure mathematics, where problems arrive pre-formulated and often demand novel techniques, statistical proof requires first modeling a scientific question into a statistical framework with appropriate assumptions, and then identifying and adapting the right strategy from a repertoire of reusable domain-specific tools. Each step requires deep expertise in both the statistical literature and the real-world context being modeled. In such work, current AI assistance does not reduce the need for human expertise; it relocates that expertise to where human decision-making matters most, such as problem formulation and verification of AI-generated results, and may raise the bar for both. These findings yield practical suggestions for how statisticians can structure AI-assisted proof workflows, and point to a broader community agenda for shared resources, better AI tools, and training the next generation of researchers. Using statistical proof as a window, our study has implications for how experts structure human-AI collaboration in technical cognitive domains more broadly.

stat.OT↗

arXiv2026

BioInsight: Multi-Agent Orchestration for Interactive Biomedical Knowledge Discovery

Biomedical deep-research systems increasingly retrieve and synthesize scientific evidence, but their outputs typically collapse heterogeneous evidence into static text, making provenance difficult to inspect and reuse. We formulate evidence-centered biomedical knowledge discovery, where disease-associated protein signals are transformed into a structured evidence state connecting proteins, pathways, publications, interactions, claims, and uncertainty. We introduce BioInsight, a provenance-preserving multi-agent orchestration framework built around typed artifact contracts and an independent Search Agent that decouples evidence acquisition from downstream mechanistic reasoning, supporting both the citation-grounded report and an interactive evidence workspace, without independently regenerating evidence for visualization. We evaluate BioInsight on standardized biomedical QA, challenging protein-function reasoning, and end-to-end biomedical evidence synthesis. The results demonstrate that BioInsight achieves better traceability and ranking performance than standard search-augmented baselines, and suggest that biomedical AI systems should move toward provenance-preserving, interactive evidence artifacts.

cs.AI↗

arXiv2026

Supervised Fine-tuning with Synthetic Rationale Data Hurts Real-World Disease Prediction

Supervised fine-tuning with synthetic rationale data is widely assumed to improve language model performance on clinical prediction tasks by teaching models not just what to predict but why. We test this assumption on five-year Alzheimer's disease and related dementias (ADRD) prediction from longitudinal health histories. Across a large-scale controlled experiment of 504 configurations, we find that rationale-based SFT consistently and substantially hurts prediction performance relative to label-only fine-tuning. The degradation persists across model families and data scales, and is not resolved by using a reasoning-oriented base model. Crucially, the failure is not explained by poor rationale quality: human expert annotation confirms that the generated rationales are medically accurate and faithfully grounded in patient-specific evidence, and few-shot experiments show that the same rationales improve performance when used as inference-time demonstrations rather than training targets. We identify the root cause as a structural conflict between narrative plausibility and discriminative optimization. We hope our work paves the path toward a more precise understanding of when and how rationale-based supervision helps and when it does not, guiding the responsible development of language models for high-stakes clinical prediction.

cs.AI↗

arXiv2026

Generalized Heterogeneous Functional Model with Applications to Large-scale Mobile Health Data

Physical activity is crucial for human health. With the increasing availability of large-scale mobile health data, strong associations have been found between physical activity and various diseases. However, accurately capturing this complex relationship is challenging, possibly because it varies across different subgroups of subjects, especially in large-scale datasets. To fill this gap, we propose a generalized heterogeneous functional method which simultaneously estimates functional effects and identifies subgroups within the generalized functional regression framework. The proposed method captures subgroup-specific functional relationships between physical activity and diseases, providing a more nuanced understanding of these associations. Additionally, we develop a pre-clustering method that enhances computational efficiency for large-scale data through a finer partition of subjects compared to true subgroups. We further introduce a testing procedure to assess whether the different subgroups exhibit distinct functional effects. In the real data application, we examine the impact of physical activity on the risk of dementia using the UK Biobank dataset, which includes over 96,433 participants. Our proposed method outperforms existing methods in future-day prediction accuracy, identifying three distinct subgroups, with detailed scientific interpretations for each subgroup. We also demonstrate the theoretical consistency of our methods. Codes implementing the proposed method are available at: https://github.com/xiaojing777/GHFM.

stat.ME↗

arXiv2025

How to Find Fantastic AI Papers: Self-Rankings as a Powerful Predictor of Scientific Impact Beyond Peer Review

Peer review in academic research aims not only to ensure factual correctness but also to identify work of high scientific potential that can shape future research directions. This task is especially critical in fast-moving fields such as artificial intelligence (AI), yet it has become increasingly difficult given the rapid growth of submissions. In this paper, we investigate an underexplored measure for identifying high-impact research: authors' own rankings of their multiple submissions to the same AI conference. Grounded in game-theoretic reasoning, we hypothesize that self-rankings are informative because authors possess unique understanding of their work's conceptual depth and long-term promise. To test this hypothesis, we conducted a large-scale experiment at a leading AI conference, where 1,342 researchers self-ranked their 2,592 submissions by perceived quality. Tracking outcomes over more than a year, we found that papers ranked highest by their authors received twice as many citations as their lowest-ranked counterparts; self-rankings were especially effective at identifying highly cited papers (those with over 150 citations). Moreover, we showed that self-rankings outperformed peer review scores in predicting future citation counts. Our results remained robust after accounting for confounders such as preprint posting time and self-citations. Together, these findings demonstrate that authors' self-rankings provide a reliable and valuable complement to peer review for identifying and elevating high-impact research in AI.

stat.AP↗

arXiv2025

DynamiCare: A Dynamic Multi-Agent Framework for Interactive and Open-Ended Medical Decision-Making

The rise of Large Language Models (LLMs) has enabled the development of specialized AI agents with domain-specific reasoning and interaction capabilities, particularly in healthcare. While recent frameworks simulate medical decision-making, they largely focus on single-turn tasks where a doctor agent receives full case information upfront -- diverging from the real-world diagnostic process, which is inherently uncertain, interactive, and iterative. In this paper, we introduce MIMIC-Patient, a structured dataset built from the MIMIC-III electronic health records (EHRs), designed to support dynamic, patient-level simulations. Building on this, we propose DynamiCare, a novel dynamic multi-agent framework that models clinical diagnosis as a multi-round, interactive loop, where a team of specialist agents iteratively queries the patient system, integrates new information, and dynamically adapts its composition and strategy. We demonstrate the feasibility and effectiveness of DynamiCare through extensive experiments, establishing the first benchmark for dynamic clinical decision-making with LLM-powered agents.

cs.AI↗

arXiv2025

High-dimensional statistical inference for linkage disequilibrium score regression and its cross-ancestry extensions

Linkage disequilibrium score regression (LDSC) has emerged as an essential tool for genetic and genomic analyses of complex traits, utilizing high-dimensional data derived from genome-wide association studies (GWAS). LDSC computes the linkage disequilibrium (LD) scores using an external reference panel, and integrates the LD scores with only summary data from the original GWAS. In this paper, we investigate LDSC within a fixed-effect data integration framework, underscoring its ability to merge multi-source GWAS data and reference panels. In particular, we take account of the genome-wide dependence among the high-dimensional GWAS summary statistics, along with the block-diagonal dependence pattern in estimated LD scores. Our analysis uncovers several key factors of both the original GWAS and reference panel datasets that determine the performance of LDSC. We show that it is relatively feasible for LDSC-based estimators to achieve asymptotic normality when applied to genome-wide genetic variants (e.g., in genetic variance and covariance estimation), whereas it becomes considerably challenging when we focus on a much smaller subset of genetic variants (e.g., in partitioned heritability analysis). Moreover, by modeling the disparities in LD patterns across different populations, we unveil that LDSC can be expanded to conduct cross-ancestry analyses using data from distinct global populations (such as European and Asian). We validate our theoretical findings through extensive numerical evaluations using real genetic data from the UK Biobank study.

stat.ME↗

arXiv2025

On self-training of summary data with genetic applications

Prediction model training is often hindered by limited access to individual-level data due to privacy concerns and logistical challenges, particularly in biomedical research. Resampling-based self-training presents a promising approach for building prediction models using only summary-level data. These methods leverage summary statistics to sample pseudo datasets for model training and parameter optimization, allowing for model development without individual-level data. Although increasingly used in precision medicine, the general behaviors of self-training remain unexplored. In this paper, we leverage a random matrix theory framework to establish the statistical properties of self-training algorithms for high-dimensional sparsity-free summary data. We demonstrate that, within a class of linear estimators, resampling-based self-training achieves the same asymptotic predictive accuracy as conventional training methods that require individual-level datasets. These results suggest that self-training with only summary data incurs no additional cost in prediction accuracy, while offering significant practical convenience. Our analysis provides several valuable insights and counterintuitive findings. For example, while pseudo-training and validation datasets are inherently dependent, their interdependence unexpectedly cancels out when calculating prediction accuracy measures, preventing overfitting in self-training algorithms. Furthermore, we extend our analysis to show that the self-training framework maintains this no-cost advantage when combining multiple methods or when jointly training on data from different distributions. We numerically validate our findings through simulations and real data analyses using the UK Biobank. Our study highlights the potential of resampling-based self-training to advance genetic risk prediction and other fields that make summary data publicly available.

stat.ME↗

arXiv2025

A semi-adaptive finite difference method for simulating two-sided fractional convection-diffusion quenching problems

This paper investigates quenching solutions of an one-dimensional, two-sided Riemann-Liouville fractional order convection-diffusion problem. Fractional order spatial derivatives are discretized using weighted averaging approximations in conjunction with standard and shifted Grünwald formulas. The advective term is handled utilizing a straightforward Euler formula, resulting in a semi-discretized system of nonlinear ordinary differential equations. The conservativeness of the proposed scheme is rigorously proved and validated through simulation experiments. The study is further advanced to a fully discretized, semi-adaptive finite difference method. Detailed analysis is implemented for the monotonicity, positivity and stability of the scheme. Investigations are carried out to assess the potential impacts of the fractional order on quenching location, quenching time, and critical length. The computational results are thoroughly discussed and analyzed, providing a more comprehensive understanding of the quenching phenomena modeled through two-sided fractional order convection-diffusion problems.

math.AP↗

arXiv2025

Generalized Heterogeneous Functional Model with Applications to Large-scale Mobile Health Data

Physical activity is crucial for human health. With the increasing availability of large-scale mobile health data, strong associations have been found between physical activity and various diseases. However, accurately capturing this complex relationship is challenging, possibly because it varies across different subgroups of subjects, especially in large-scale datasets. To fill this gap, we propose a generalized heterogeneous functional method which simultaneously estimates functional effects and identifies subgroups within the generalized functional regression framework. The proposed method captures subgroup-specific functional relationships between physical activity and diseases, providing a more nuanced understanding of these associations. Additionally, we introduce a pre-clustering method that enhances computational efficiency for large-scale data through a finer partition of subjects compared to true subgroups. In the real data application, we examine the impact of physical activity on the risk of mental disorders and Parkinson's disease using the UK Biobank dataset, which includes over 79,000 participants. Our proposed method outperforms existing methods in future-day prediction accuracy, identifying four subgroups for mental disorder outcomes and three subgroups for Parkinson's disease diagnosis, with detailed scientific interpretations for each subgroup. We also demonstrate theoretical consistency of our methods. Supplementary materials are available online. Codes implementing the proposed method are available at: https://github.com/xiaojing777/GHFM.

stat.AP↗

arXiv2024

Uncertainty of high-dimensional genetic data prediction with polygenic risk scores

In many predictive tasks, there are a large number of true predictors with weak signals, leading to substantial uncertainties in prediction outcomes. The polygenic risk score (PRS) is an example of such a scenario, where many genetic variants are used as predictors for complex traits, each contributing only a small amount of information. Although PRS has been a standard tool in genetic predictions, its uncertainty remains largely unexplored. In this paper, we aim to establish the asymptotic normality of PRS in high-dimensional predictions without sparsity constraints. We investigate the popular marginal and ridge-type estimators in PRS applications, developing central limit theorems for both individual-level predicted values (e.g., genetically predicted human height) and cohort-level prediction accuracy measures (e.g., overall predictive $R$-squared in the testing dataset). Our results demonstrate that ignoring the prediction-induced uncertainty can lead to substantial underestimation of the true variance of PRS-based estimators, which in turn may cause overconfidence in the accuracy of confidence intervals and hypothesis testing. These findings provide key insights omitted by existing first-order asymptotic studies of high-dimensional sparsity-free predictions, which often focus solely on the point limits of predictive risks. We develop novel and flexible second-order random matrix theory results to assess the asymptotic normality of functionals with a general covariance matrix, without assuming Gaussian distributions for the data. We evaluate our theoretical results through extensive numerical analyses using real data from the UK Biobank. Our analysis underscores the importance of incorporating uncertainty assessments at both the individual and cohort levels when applying and interpreting PRS.

stat.ME↗

arXiv2024

The Exact Risks of Reference Panel-based Regularized Estimators

Reference panel-based estimators have become widely used in genetic prediction of complex traits due to their ability to address data privacy concerns and reduce computational and communication costs. These estimators estimate the covariance matrix of predictors using an external reference panel, instead of relying solely on the original training data. In this paper, we investigate the performance of reference panel-based $L_1$ and $L_2$ regularized estimators within a unified framework based on approximate message passing (AMP). We uncover several key factors that influence the accuracy of reference panel-based estimators, including the sample sizes of the training data and reference panels, the signal-to-noise ratio, the underlying sparsity of the signal, and the covariance matrix among predictors. Our findings reveal that, even when the sample size of the reference panel matches that of the training data, reference panel-based estimators tend to exhibit lower accuracy compared to traditional regularized estimators. Furthermore, we observe that this performance gap widens as the amount of training data increases, highlighting the importance of constructing large-scale reference panels to mitigate this issue. To support our theoretical analysis, we develop a novel non-separable matrix AMP framework capable of handling the complexities introduced by a general covariance matrix and the additional randomness associated with a reference panel. We validate our theoretical results through extensive simulation studies and real data analyses using the UK Biobank database.

stat.ME↗

arXiv2023

Minimax rates of convergence for nonparametric location-scale models

This paper studies minimax rates of convergence for nonparametric location-scale models, which include mean, quantile and expectile regression settings. Under Hellinger differentiability on the error distribution and other mild conditions, we show that the minimax rate of convergence for estimating the regression function under the squared $L_2$ loss is determined by the metric entropy of the nonparametric function class. Different error distributions, including asymmetric Laplace distribution, asymmetric connected double truncated gamma distribution, connected normal-Laplace distribution, Cauchy distribution and asymmetric normal distribution are studied as examples. Applications on low order interaction models and multiple index models are also given.

math.ST↗

arXiv2022

Statistical learning methods for neuroimaging data analysis with applications

The aim of this paper is to provide a comprehensive review of statistical challenges in neuroimaging data analysis from neuroimaging techniques to large-scale neuroimaging studies to statistical learning methods. We briefly review eight popular neuroimaging techniques and their potential applications in neuroscience research and clinical translation. We delineate the four common themes of neuroimaging data and review major image processing analysis methods for processing neuroimaging data at the individual level. We briefly review four large-scale neuroimaging-related studies and a consortium on imaging genomics and discuss four common themes of neuroimaging data analysis at the population level. We review nine major population-based statistical analysis methods and their associated statistical challenges and present recent progress in statistical methodology to address these challenges.

stat.AP↗

arXiv2022

Combining Predictions of Auto Insurance Claims

This paper aims to better predict highly skewed auto insurance claims by combining candidate predictions. We analyze a version of the Kangaroo Auto Insurance company data and study the effects of combining different methods using five measures of prediction accuracy. The results show the following. First, when there is an outstanding (in terms of Gini Index) prediction among the candidates, the "forecast combination puzzle" phenomenon disappears. The simple average method performs much worse than the more sophisticated model combination methods, indicating that combining different methods could help us avoid performance degradation. Second, the choice of the prediction accuracy measure is crucial in defining the best candidate prediction for "low frequency and high severity" (LFHS) data. For example, mean square error (MSE) does not distinguish well between model combination methods, as the values are close. Third, the performances of different model combination methods can differ drastically. We propose using a new model combination method, named ARM-Tweedie, for such LFHS data; it benefits from an optimal rate of convergence and exhibits a desirable performance in several measures for the Kangaroo data. Fourth, overall, model combination methods improve the prediction accuracy for auto insurance claim costs. In particular, Adaptive Regression by Mixing (ARM), ARM-Tweedie, and constrained Linear Regression can improve forecast performance when there are only weak learners or when no dominant learner exists.

stat.AP↗

arXiv2022

Estimating trans-ancestry genetic correlation with unbalanced data resources

The aim of this paper is to propose a novel estimation method of using genetic-predicted observations to estimate trans-ancestry genetic correlations, which describes how genetic architecture of complex traits varies among populations, in genome-wide association studies (GWAS). Our new estimator corrects for prediction errors caused by high-dimensional weak GWAS signals, while addressing the heterogeneity of GWAS data across ethnicities, such as linkage disequilibrium (LD) differences, which can lead to biased findings in homogeneity-agnostic analyses. Moreover, our estimator only requires one population to have a large GWAS sample size, and the second population can only have a much smaller number of participants (for example, hundreds). It is designed to specifically address the unbalanced data resources such that the GWAS sample size for European populations is usually larger than that of non-European ancestry groups. Extensive simulations and real data analyses of 30 complex traits in the UK Biobank study show that our method is capable of providing reliable estimates of a wide range of complex traits. Our results provide deep insights into the transferability of population-specific genetic findings.

stat.ME↗

arXiv2022

On block-wise and reference panel-based estimators for genetic data prediction in high dimensions

Genetic prediction of complex traits and diseases has attracted enormous attention in precision medicine, mainly because it has the potential to translate discoveries from genome-wide association studies (GWAS) into medical advances. As the high dimensional covariance matrix (or the linkage disequilibrium (LD) pattern) of genetic variants has a block-diagonal structure, many existing methods attempt to account for the dependence among variants in predetermined local LD blocks/regions. Moreover, due to privacy restrictions and data protection concerns, genetic variant dependence in each LD block is typically estimated from external reference panels rather than the original training dataset. This paper presents a unified analysis of block-wise and reference panel-based estimators in a high-dimensional prediction framework without sparsity restrictions. We find that, surprisingly, even when the covariance matrix has a block-diagonal structure with well-defined boundaries, block-wise estimation methods adjusting for local dependence can be substantially less accurate than methods controlling for the whole covariance matrix. Further, estimation methods built on the original training dataset and external reference panels are likely to have varying performance in high dimensions, which may reflect the cost of having only access to summary level data from the training dataset. This analysis is based on our novel results in random matrix theory for block-diagonal covariance matrix. We numerically evaluate our results using extensive simulations and the large-scale UK Biobank real data analysis of 36 complex traits.

stat.ME↗