arXiv subjects

Peng Wei

Publications and source records attributed to Peng Wei.

At least 55 records · Page 3Linked to original sources

arXiv2026

A Robust Geometric Distortion Solution for Main Survey Camera of CSST

The advancement in sensitivity and field of view of next-generation wide-field survey telescopes requires astrometric measurements with high precision, even in the presence of significant geometric distortions. To address this challenge, we develop a Weighted Polynomial Distortion Correction in 2-Phase (WPDC-2P) method. This approach enhances stellar cross-matching, incorporates distance-based weighting into the traditional polynomial fitting, and employs a look-up table to absorb the remaining distortion residuals. Validated on simulated data from the Main Survey Camera of the \emph{Chinese Space Station Survey Telescope} (CSST), incorporating geometric distortions up to approximately $200$ pixels, the method achieves astrometric standard deviation ranging from 0.013 to 0.107 pixels (0.03 pixels for the $g$-1 detector) across all 18 detectors. Under extreme crowding conditions (e.g., globular cluster NGC 2298), the astrometric precision for the $g$-1 detector reaches 0.05-pixel level within the central region ($r_d < 4000$), despite a centroiding precision of $\sim$0.04 pixels. When applied to the Beijing-Arizona Sky Survey data, for which the standard pipeline delivers an astrometric uncertainty of $\sim$20 mas, our method reduces the positional scatter to $ σ_{Δα}=5.494$ mas (0.01 pixels) and $ σ_{Δδ}=9.981$ mas (0.02 pixels) using only a weighted 3rd-order polynomial correction. The method has been integrated into the CSST data processing pipeline and is prepared for further refinement using on-orbit calibration data.

astro-ph.IM↗

arXiv2026

Training a Large Language Model for Medical Coding Using Privacy-Preserving Synthetic Clinical Data

Improving the accuracy and reliability of medical coding reduces clinician burnout and supports revenue cycle processes, freeing providers to focus more on patient care. However, automating the assignment of ICD-10-CM and CPT codes from clinical documentation remains a challenge due to heterogeneous records, nuanced coding guidelines, and long-tail distributions. Large language models have been proposed to help or automate specific medical coding tasks. However, foundation models are not explicitly trained for medical coding and zero-shot coding has yielded poor results. We investigate whether a modern open-weight foundation model can be adapted for an expert-level medical coding task using privacy-preserving synthetic training data derived from electronic health records. We fine-tune Llama 3-70B on pairs of clinical notes and gold codes generated from EHR-grounded templates and coding policies, then evaluate exact-code prediction for ICD-10-CM and CPT. A zero-shot baseline with the unadapted model achieved an F1 score of 0.18 for exact code match. After fine-tuning on the synthetic corpus, exact-match F1 exceeded 0.70, representing a large absolute gain across both code systems. Notably, performance remained high on complex categories that often require multi-step clinical reasoning and code composition, including Advanced Illness and Frailty classes, and the model retained its performance on medical comprehension tasks. These results indicate that synthetic, policy-aware data can efficiently teach a general-purpose large language model to support precise medical coding without exposing protected health information. The approach offers a practical path for training coding agents safely and iteratively on specific tasks that represent real-world populations.

cs.CL↗

arXiv2026

Selecting Optimal Stellar Calibration Fields for the CSST Imaging Survey

The Chinese Space Station Survey Telescope (CSST) will perform a decade-long high-precision wide-field imaging survey that relies on rigorous on-orbit calibration. This necessitates stable celestial benchmark fields to maintain photometric and astrometric consistency throughout the mission lifetime. We establish comprehensive selection criteria including observational visibility, stellar number density, bright-star contamination, and interstellar dust extinction. Using the CSST Observation Strategy Analysis Tool (COSAT) and all-sky dust maps from Planck and SFD, we constrain eligible regions to the ranges of ecliptic latitude $ |β| > 50^\circ$ and galactic latitude $|b| > 15^\circ$. From an initial sample of 29 candidate clusters meeting these spatial constraints, six globular clusters (M13, M92, NGC 104, NGC 362, NGC 1261, and NGC 1851) are identified as optimal calibration fields, fulfilling all the critical criteria. These selected clusters are recommended as optimal calibration field candidates for CSST's on-orbit calibration program, and are fundamental to achieving unprecedented photometric precision in CSST's space-based survey.

astro-ph.SR↗

arXiv2026

DomainCQA: Crafting Knowledge-Intensive QA from Domain-Specific Charts

Chart Question Answering (CQA) evaluates Multimodal Large Language Models (MLLMs) on visual understanding and reasoning over chart data. However, existing benchmarks mostly test surface-level parsing, such as reading labels and legends, while overlooking deeper scientific reasoning. We propose DomainCQA, a framework for constructing domain-specific CQA benchmarks that emphasize both visual comprehension and knowledge-intensive reasoning. It integrates complexity-aware chart selection, multitier QA generation, and expert validation. Applied to astronomy, DomainCQA yields AstroChart, a benchmark of 1,690 QA pairs over 482 charts, exposing persistent weaknesses in fine-grained perception, numerical reasoning, and domain knowledge integration across 21 MLLMs. Fine-tuning on AstroChart improves performance across fundamental and advanced tasks. Pilot QA sets in biochemistry, economics, medicine, and social science further demonstrate DomainCQA's generality. Together, our results establish DomainCQA as a unified pipeline for constructing and augmenting domain-specific chart reasoning benchmarks.

cs.CL↗

arXiv2026

A Survey of Security Challenges and Solutions for Advanced Air Mobility and eVTOL Aircraft

This survey reviews the existing and envisioned security vulnerabilities and defense mechanisms relevant to Advanced Air Mobility (AAM) systems, with a focus on electric vertical takeoff and landing (eVTOL) aircraft. Drawing from vulnerabilities in the avionics in commercial aviation and the automated unmanned aerial systems (UAS), the paper presents a taxonomy of attacks, analyzes mitigation strategies, and proposes a secure system architecture tailored to the future AAM ecosystem. The paper also highlights key threat vectors, including Global Positioning System (GPS) jamming/spoofing, ATC radio frequency misuse, attacks on TCAS and ADS-B, possible backdoor via Electronic Flight Bag (EFB), new vulnerabilities introduced by aircraft automation and connectivity, and risks from flight management system (FMS) software, database and cloud services. Finally, this paper describes emerging defense techniques against these attacks, and open technical problems to address toward better defense mechanisms.

cs.CR↗

arXiv2026

A Survey of Security Challenges and Solutions for UAS Traffic Management (UTM) and small Unmanned Aerial Systems (sUAS)

The rapid growth of small Unmanned Aerial Systems (sUAS) for civil and commercial missions has intensified concerns about their resilience to cyber-security threats. Operating within the emerging UAS Traffic Management (UTM) framework, these lightweight and highly networked platforms depend on secure communication, navigation, and surveillance (CNS) subsystems that are vulnerable to spoofing, jamming, hijacking, and data manipulation. While prior reviews of UAS security addressed these challenges at a conceptual level, a detailed, system-oriented analysis for resource-constrained sUAS remains lacking. This paper presents a comprehensive survey of cyber-security vulnerabilities and defenses tailored to the sUAS and UTM ecosystem. We organize existing research across the full cyber-physical stack, encompassing CNS, data links, sensing and perception, UTM cloud access, and software integrity layers, and classify attack vectors according to their technical targets and operational impacts. Correspondingly, we review defense mechanisms ranging from classical encryption and authentication to adaptive intrusion detection, lightweight cryptography, and secure firmware management. By mapping threats to mitigation strategies and evaluating their scalability and practical effectiveness, this work establishes a unified taxonomy and identifies open challenges for achieving safe, secure, and scalable sUAS operations within future UTM environments.

cs.CR↗

arXiv2026

MedDialogRubrics: A Comprehensive Benchmark and Evaluation Framework for Multi-turn Medical Consultations in Large Language Models

Medical conversational AI (AI) plays a pivotal role in the development of safer and more effective medical dialogue systems. However, existing benchmarks and evaluation frameworks for assessing the information-gathering and diagnostic reasoning abilities of medical large language models (LLMs) have not been rigorously evaluated. To address these gaps, we present MedDialogRubrics, a novel benchmark comprising 5,200 synthetically constructed patient cases and over 60,000 fine-grained evaluation rubrics generated by LLMs and subsequently refined by clinical experts, specifically designed to assess the multi-turn diagnostic capabilities of LLM. Our framework employs a multi-agent system to synthesize realistic patient records and chief complaints from underlying disease knowledge without accessing real-world electronic health records, thereby mitigating privacy and data-governance concerns. We design a robust Patient Agent that is limited to a set of atomic medical facts and augmented with a dynamic guidance mechanism that continuously detects and corrects hallucinations throughout the dialogue, ensuring internal coherence and clinical plausibility of the simulated cases. Furthermore, we propose a structured LLM-based and expert-annotated rubric-generation pipeline that retrieves Evidence-Based Medicine (EBM) guidelines and utilizes the reject sampling to derive a prioritized set of rubric items ("must-ask" items) for each case. We perform a comprehensive evaluation of state-of-the-art models and demonstrate that, across multiple assessment dimensions, current models face substantial challenges. Our results indicate that improving medical dialogue will require advances in dialogue management architectures, not just incremental tuning of the base-model.

cs.CL↗

arXiv2026

Transformer-based Multi-agent Reinforcement Learning for Separation Assurance in Structured and Unstructured Airspaces

Conventional optimization-based metering depends on strict adherence to precomputed schedules, which limits the flexibility required for the stochastic operations of Advanced Air Mobility (AAM). In contrast, multi-agent reinforcement learning (MARL) offers a decentralized, adaptive framework that can better handle uncertainty, required for safe aircraft separation assurance. Despite this advantage, current MARL approaches often overfit to specific airspace structures, limiting their adaptability to new configurations. To improve generalization, we recast the MARL problem in a relative polar state space and train a transformer encoder model across diverse traffic patterns and intersection angles. The learned model provides speed advisories to resolve conflicts while maintaining aircraft near their desired cruising speeds. In our experiments, we evaluated encoder depths of 1, 2, and 3 layers in both structured and unstructured airspaces, and found that a single encoder configuration outperformed deeper variants, yielding near-zero near mid-air collision rates and shorter loss-of-separation infringements than the deeper configurations. Additionally, we showed that the same configuration outperforms a baseline model designed purely with attention. Together, our results suggest that the newly formulated state representation, novel design of neural network architecture, and proposed training strategy provide an adaptable and scalable decentralized solution for aircraft separation assurance in both structured and unstructured airspaces.

cs.RO↗

arXiv2026

Unsupervised dense random survival forests identify interpretable patient profiles with heterogeneous treatment benefit

Precision oncology aims to prescribe the optimal cancer treatment to the right patients, maximizing therapeutic benefits. However, identifying patient subgroups that may benefit more from experimental cancer treatments based on randomized clinical trials presents a significant analytical challenge. To address this, we introduce a novel unsupervised machine learning approach based on very dense random survival forests (up to 100,000 trees), equipped with a new splitting rule that explicitly targets treatment-effect heterogeneity. This method is robust, interpretable, and effectively identifies responsive subgroups. Extensive simulations confirm its ability to detect heterogeneous patient responses and distinguish between datasets with and without heterogeneity, while maintaining a stringent Type I error rate of 1%. We further validate its performance using Phase III randomized clinical trial datasets, demonstrating significant patient heterogeneity in treatment response based on baseline characteristics.

stat.ME↗

arXiv2025

GAPS: A Clinically Grounded, Automated Benchmark for Evaluating AI Clinicians

Current benchmarks for AI clinician systems, often based on multiple-choice exams or manual rubrics, fail to capture the depth, robustness, and safety required for real-world clinical practice. To address this, we introduce the GAPS framework, a multidimensional paradigm for evaluating Grounding (cognitive depth), Adequacy (answer completeness), Perturbation (robustness), and Safety. Critically, we developed a fully automated, guideline-anchored pipeline to construct a GAPS-aligned benchmark end-to-end, overcoming the scalability and subjectivity limitations of prior work. Our pipeline assembles an evidence neighborhood, creates dual graph and tree representations, and automatically generates questions across G-levels. Rubrics are synthesized by a DeepResearch agent that mimics GRADE-consistent, PICO-driven evidence review in a ReAct loop. Scoring is performed by an ensemble of large language model (LLM) judges. Validation confirmed our automated questions are high-quality and align with clinician judgment (90% agreement, Cohen's Kappa 0.77). Evaluating state-of-the-art models on the benchmark revealed key failure modes: performance degrades sharply with increased reasoning depth (G-axis), models struggle with answer completeness (A-axis), and they are highly vulnerable to adversarial perturbations (P-axis) as well as certain safety issues (S-axis). This automated, clinically-grounded approach provides a reproducible and scalable method for rigorously evaluating AI clinician systems and guiding their development toward safer, more reliable clinical practice. The benchmark dataset GAPS-NSCLC-preview and evaluation code are publicly available at https://huggingface.co/datasets/AQ-MedAI/GAPS-NSCLC-preview and https://github.com/AQ-MedAI/MedicalAiBenchEval.

cs.CL↗

arXiv2025

Multi-Agent Deep Research: Training Multi-Agent Systems with M-GRPO

Multi-agent systems perform well on general reasoning tasks. However, the lack of training in specialized areas hinders their accuracy. Current training methods train a unified large language model (LLM) for all agents in the system. This may limit the performances due to different distributions underlying for different agents. Therefore, training multi-agent systems with distinct LLMs should be the next step to solve. However, this approach introduces optimization challenges. For example, agents operate at different frequencies, rollouts involve varying sub-agent invocations, and agents are often deployed across separate servers, disrupting end-to-end gradient flow. To address these issues, we propose M-GRPO, a hierarchical extension of Group Relative Policy Optimization designed for vertical Multi-agent systems with a main agent (planner) and multiple sub-agents (multi-turn tool executors). M-GRPO computes group-relative advantages for both main and sub-agents, maintaining hierarchical credit assignment. It also introduces a trajectory-alignment scheme that generates fixed-size batches despite variable sub-agent invocations. We deploy a decoupled training pipeline in which agents run on separate servers and exchange minimal statistics via a shared store. This enables scalable training without cross-server backpropagation. In experiments on real-world benchmarks (e.g., GAIA, XBench-DeepSearch, and WebWalkerQA), M-GRPO consistently outperforms both single-agent GRPO and multi-agent GRPO with frozen sub-agents, demonstrating improved stability and sample efficiency. These results show that aligning heterogeneous trajectories and decoupling optimization across specialized agents enhances tool-augmented reasoning tasks.

cs.AI↗

arXiv2025

Mock Observations for the CSST Mission: Main Surveys-the Slitless Spectroscopy Simulation

The China Space Station Telescope (CSST), slated to become China's largest space-based optical telescope in the coming decade, is designed to conduct wide-field sky surveys with high spatial resolution. Among its key observational modes, slitless spectral observation allows simultaneous imaging and spectral data acquisition over a wide field of view, offering significant advantages for astrophysical studies. Currently, the CSST is in the development phase and lacks real observational data. As a result, the development of its data processing pipeline and scientific pre-research must rely on the mock data generated through simulations. This work focuses on developing a simulation framework for the CSST slitless spectral imaging system, analyzing its spectral dispersing properties and structural design. Additionally, the detection performance of the slitless spectral system is assessed for various astrophysical targets. Simulation results demonstrate that nearly all 1st order spectra are accompanied by corresponding 0th order images, facilitating accurate source identification. Furthermore, the GI spectral band exhibits superior detection efficiency compared to the GV and GU bands, establishing it as the primary observational band for stellar and galactic studies. This work successfully develops a simulation framework for the CSST slitless spectroscopic equipment.

astro-ph.IM↗

arXiv2025

Mock Observations for the CSST Mission: End-to-End Performance Modeling of Optical System

This study presents a comprehensive end-to-end simulation analysis of the optical imaging performance of the China Survey Space Telescope (CSST) under in-orbit conditions. An integrated system model incorporating five static and two dynamic error sub-models was established. Wavefront errors were calculated for each sub-model and compared to the integrated system error to quantify the individual contributions to image degradation. At the detector level, wavefront error, point spread function (PSF), and ellipticity were evaluated across the full field of view (FOV). The average radius of 80\% encircled energy (REE80) of the PSF under full-error conditions was determined for 25 field points, yielding a value of 0.114 arcseconds. Furthermore, the calculations indicate a correlation between the wavefront distribution and the ellipticity distribution within the optical system. By optimizing the wavefront distribution, it is possible to adjust the ellipticity distribution of the PSF across the full FOV. The end-to-end simulation approach adopted in this paper provides a theoretical foundation for improving the image quality in large-aperture, off-axis space telescopes.

astro-ph.IM↗

arXiv2025

Self-Rewarding Rubric-Based Reinforcement Learning for Open-Ended Reasoning

Open-ended evaluation is essential for deploying large language models in real-world settings. In studying HealthBench, we observe that using the model itself as a grader and generating rubric-based reward signals substantially improves reasoning performance. Remarkably, the trained model also becomes a stronger grader. Motivated by this, we introduce Self-Rewarding Rubric-Based Reinforcement Learning for Open-Ended Reasoning, a lightweight framework that enables faster and more resource-efficient training while surpassing baselines. Remarkably, on Qwen3-32B, training with just the 4000-sample HealthBench Easy subset is sufficient to obtain a model that exceeds GPT-5 on HealthBench Hard. Incorporating a small amount of teacher-graded data further enhances performance for less capable models.

cs.CL↗

arXiv2025

HIRAG: Hierarchical-Thought Instruction-Tuning Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) has become a fundamental paradigm for addressing the challenges faced by large language models in handling real-time information and domain-specific problems. Traditional RAG systems primarily rely on the in-context learning (ICL) capabilities of the large language model itself. Still, in-depth research on the specific capabilities needed by the RAG generation model is lacking, leading to challenges with inconsistent document quality and retrieval system imperfections. Even the limited studies that fine-tune RAG generative models often \textit{lack a granular focus on RAG task} or \textit{a deeper utilization of chain-of-thought processes}. To address this, we propose that RAG models should possess three progressively hierarchical abilities (1) Filtering: the ability to select relevant information; (2) Combination: the ability to combine semantic information across paragraphs; and (3) RAG-specific reasoning: the ability to further process external knowledge using internal knowledge. Thus, we introduce our new RAG instruction fine-tuning method, Hierarchical-Thought Instruction-Tuning Retrieval-Augmented Generation (HIRAG) incorporates a "think before answering" strategy. This method enhances the model's open-book examination capability by utilizing multi-level progressive chain-of-thought. Experiments show that the HIRAG training strategy significantly improves the model's performance on datasets such as RGB, PopQA, MuSiQue, HotpotQA, and PubmedQA.

cs.CL↗

arXiv2025

HANRAG: Heuristic Accurate Noise-resistant Retrieval-Augmented Generation for Multi-hop Question Answering

The Retrieval-Augmented Generation (RAG) approach enhances question-answering systems and dialogue generation tasks by integrating information retrieval (IR) technologies with large language models (LLMs). This strategy, which retrieves information from external knowledge bases to bolster the response capabilities of generative models, has achieved certain successes. However, current RAG methods still face numerous challenges when dealing with multi-hop queries. For instance, some approaches overly rely on iterative retrieval, wasting too many retrieval steps on compound queries. Additionally, using the original complex query for retrieval may fail to capture content relevant to specific sub-queries, resulting in noisy retrieved content. If the noise is not managed, it can lead to the problem of noise accumulation. To address these issues, we introduce HANRAG, a novel heuristic-based framework designed to efficiently tackle problems of varying complexity. Driven by a powerful revelator, HANRAG routes queries, decomposes them into sub-queries, and filters noise from retrieved documents. This enhances the system's adaptability and noise resistance, making it highly capable of handling diverse queries. We compare the proposed framework against other leading industry methods across various benchmarks. The results demonstrate that our framework obtains superior performance in both single-hop and multi-hop question-answering tasks.

cs.CL↗

arXiv2025

MedResearcher-R1: Expert-Level Medical Deep Researcher via A Knowledge-Informed Trajectory Synthesis Framework

Recent developments in Large Language Model (LLM)-based agents have shown impressive capabilities spanning multiple domains, exemplified by deep research systems that demonstrate superior performance on complex information-seeking and synthesis tasks. While general-purpose deep research agents have shown impressive capabilities, they struggle significantly with medical domain challenges, as evidenced by leading proprietary systems achieving limited accuracy on complex medical benchmarks. The key limitations are: (1) the model lacks sufficient dense medical knowledge for clinical reasoning, and (2) the framework is constrained by the absence of specialized retrieval tools tailored for medical contexts. We present a medical deep research agent that addresses these challenges through two core innovations. First, we develop a novel data synthesis framework using medical knowledge graphs, extracting the longest chains from subgraphs around rare medical entities to generate complex multi-hop question-answer pairs. Second, we integrate a custom-built private medical retrieval engine alongside general-purpose tools, enabling accurate medical information synthesis. Our approach generates 2100+ diverse trajectories across 12 medical specialties, each averaging 4.2 tool interactions. Through a two-stage training paradigm combining supervised fine-tuning and online reinforcement learning with composite rewards, our MedResearcher-R1-32B model demonstrates exceptional performance, establishing new state-of-the-art results on medical benchmarks while maintaining competitive performance on general deep research tasks. Our work demonstrates that strategic domain-specific innovations in architecture, tool design, and training data construction can enable smaller open-source models to outperform much larger proprietary systems in specialized domains.

cs.CL↗

arXiv2025

Toward High-Precision Astrometry with CSST Using Multi-Gaussian Fitting of PSF

The Chinese Space Station Survey Telescope (CSST) presents significant potential for high-precision astrometry. In this study, we show that the point spread function (PSF) modeled by the discrete PSF with Multi-Gaussian function can effectively enhance the astrometric accuracy. We determine that the PSF profile can be accurately modeled by three Gaussians, which takes advantage of reduced computational complexity in PSF convolution. In sparse star fields, the lowest centering accuracy we obtain after aberration correction can be below 1 mas. We find that the proper motion errors remain below 1.0 mas/yr for point sources with five observations and approximately 0.8 mas/yr for seven observations with a time baseline of around 3.5 years. We finally demonstrate that the precision of our position measurements for stars fainter than 21 mag in the simulated CSST crowded field is better than the results from both SExtractor and DOLPHOT.

astro-ph.IM↗