Searcharxiv⌕ Search

arXiv subjects

Yikun Li

Publications and source records attributed to Yikun Li.

At least 19 recordsLinked to original sources

Minimaxity in Robust Nonparametric Regression: from Heavy-Tailed to Sub-Gaussian Assumptions

We study nonparametric regression over star-shaped function classes under strong adversarial contamination. The clean (original) observations $(\tilde{X} _{i},\tilde{Y} _{i})_{i=1} ^{n}$ are i.i.d. and satisfy $\tilde{Y} _{i}=f ^{*}(\tilde{X} _{i})+ξ_{i}$, where $f ^{*}\in \mathcal{F}$, $\tilde{X}$ is distributed as $\mathbb{P} _{X}$, and the noise is conditionally centered. An adversary may inspect the clean sample and replace at most $\lfloor εn\rfloor$ observation pairs. We assume that $\mathcal{F}\subset L _{2}(\mathbb{P} _{X})$ is star-shaped with finite diameter $d$. Under a Bernstein-type moment condition on function increments and a bounded conditional second moment for the noise, we establish the matching lower and upper bounds of order $\left((η^{*}) ^{2} \vee ε\right)\wedge d ^{2}$, uniformly over admissible star-shaped function classes, therefore characterizing the minimax rate of estimation error, where \begin{equation*} η^{*}:=\sup\left\{η>0: nη^2\leq\log \mathcal M _{\mathcal{F}}^{\text{loc}}(η,c)\right\}. \end{equation*} Here $\log \mathcal{M} _{\mathcal{F}} ^{\text{loc}}(η,c)$ denotes the local metric entropy of $\mathcal{F}$ in $L _2(\mathbb{P} _{X})$, and $c$ is a sufficiently large constant. Under a uniform relative sub-Gaussian condition on function increments and noise, the lower bound is adjusted to $\left((η^{*})^{2}\vee ε^{2}\right)\wedge d ^{2}$ while the upper bound improves to $\left((η^{*})^{2}\vee ε^{2}\log ^{2}\left(\frac{e}{ε}\right)\right)\wedge d ^{2}$, yielding minimax optimality up to logarithmic factors.

math.ST↗

AutoVend: Benchmarking LLM Agents on Long-Horizon Autonomous Vending Operations

Large language model (LLM) agents are extending agent-mediated commerce on the Web from shopping for consumers to operating businesses for sellers. In online-to-offline (O2O) commerce, decisions are issued through Web-based platforms but realized through costly physical execution. Evaluating such agents requires going beyond individual transactions to ask whether decisions made online turn into sustained economic returns. We introduce AutoVend, a benchmark in which one LLM agent manages 20 vending machines across five site types through a Web-based operating interface for 180 simulated days. Its data-grounded simulator combines a shared warehouse, procurement lead times, paid replenishment, inventory expiration, and recurring operating costs. We evaluate five LLM agents against a rule-based inventory-control baseline and diagnose their site coordination, execution, and performance over time. Four of five evaluated LLM agents fall short of this baseline, and returns per dollar of inference cost rank models differently from returns alone. Operational analyses show that gains in individual metrics do not consistently yield stronger economic growth, as inventory and execution costs erode these gains. Trajectory analysis further reveals difficulties in grounding feasible actions, managing inventory commitments, and converting feedback into effective policy revisions.

cs.AI↗

Spark-Audio 1.0 Technical Report

This paper introduces Spark-Audio 1.0, a speech foundation model trained on domestically produced computing infrastructure. The model accepts text and speech inputs and generates text outputs. It supports multilingual (99 languages) and multi-dialect (202 dialects) speech recognition, speech translation, audio perception and understanding, and both text- and speech-based dialogue, marking a shift from merely ``hearing clearly'' to genuinely ``understanding'' speech. Architecturally, Spark-Audio 1.0 couples a 0.65B-parameter dense audio encoder with a 30B-A3B Mixture-of-Experts (MoE) large language model (LLM). The audio encoder is pretrained with representation-enhanced BEST-RQ and then refined through multi-task supervised training. The audio--LLM is trained through staged pretraining, followed by post-training that combines supervised fine-tuning with on-policy distillation. In total, the model is trained on 13 million hours of audio data together with large-scale text data. We evaluate Spark-Audio 1.0 on mainstream audio and text benchmarks, as well as on in-house test sets drawn from real-world application scenarios. Despite using substantially less audio training data than Qwen3.5-Omni-Flash, Spark-Audio 1.0 achieves comparable performance on most audio and text tasks, outperforms it in multilingual speech recognition, spoken dialogue, and real-world application scenarios, and approaches the performance of larger closed-source models. Nevertheless, there remains room for improvement in instruction following and complex audio reasoning. Finally, the successful training of Spark-Audio 1.0 on domestic hardware, including Huawei Ascend 910B and Hygon BW1000 chips, demonstrates the feasibility of a domestic technical pathway for building speech foundation models.

eess.AS↗

VulContextBench: A Benchmark for Security Context Retrieval in Coding Agents

Vulnerability-detection benchmarks score the verdict an agent reaches, not the evidence it gathered. A model that recalls a CVE from pretraining therefore scores the same as one that traced the data flow. We study a task where this difference matters, deciding whether a commit introduces a vulnerability. Instead of scoring the verdict, we score whether the agent retrieved the code its conclusion depends on. We present VulContextBench, a benchmark of 111 vulnerability-introducing commits (VICs) across 83 repositories, 63 CWEs, and five languages. Existing datasets label such commits by tracing a fix back through the version history, which often points to the wrong commit. We therefore audit every case by hand against an explicit four-criterion definition of a VIC, so the benchmark does not inherit that label noise. Each case is annotated with gold context, 464 code blocks in total, each tagged by its role in the evidence for the vulnerability. We evaluate seven frontier models with precision, recall and F1 at three granularities (file, block, and line), scored separately on the context an agent viewed while exploring and on the context it finally declared as evidence. The gap between the two is the main finding. Every model opens most of the gold context while exploring, but reports only part of it as evidence. At the level of code blocks, the share of the gold context a model reports is 37 to 73 percentage points below the share it viewed. Qwen3-Coder-Next views 86.3% of the lines in annotated code blocks but cites only 12.9% in its final report. GPT-5.5, which cites the most, views 73% and reports 36%. These results highlight a gap between finding relevant code and selecting it for the final report, which verdict-level benchmarks cannot reveal.

cs.CR↗

TitanCA: Lessons from Orchestrating LLM Agents to Discover 100+ CVEs

Software vulnerabilities remain one of the most persistent threats to modern digital infrastructure. While static application security testing (SAST) tools have long served as the first line of defense, they suffer from high false-positive rates. This article presents TitanCA, a collaborative project between Singapore Management University and GovTech Singapore that orchestrates multiple large language model (LLM)-powered agents into a unified vulnerability discovery pipeline. Applied in open-source software, TitanCA has discovered 203 confirmed zero-day vulnerabilities and yielded 118 CVEs. We describe the four-module architecture, i.e., matching, filtering, inspection, and adaptation, and share key lessons from building and deploying an LLM-based vulnerability discovery solution in practice.

cs.CR↗

CoSA: Context-Aware Severity Assessment via Context Analysis with Large Language Models

Accurate vulnerability severity assessment is essential for prioritizing remediation, yet manually assessing Common Vulnerability Scoring System (CVSS) base metrics remains labor-intensive. Existing automated approaches often fail to capture the repository-level evidence required for assessing many CVSS base metrics. Such repository-aware assessment is challenging because relevant evidence is scattered across the entire repository under heavy noise. To address these challenges, we present CoSA, a Context-aware vulnerability Severity Assessment approach that infers CVSS base metrics from repository artifacts. CoSA constructs a code property graph (CPG) and applies a two-stage repository-pruning strategy: lightweight static pruning to preserve structurally proximal context, followed by an agentic large language model (LLM)-guided pruning step to retain CVSS-relevant context while collecting supporting evidence. The LLM then consolidates the retrieved repository context into compact, CVSS metric-wise textual summaries, which are fed into a lightweight transformer predictor. We also construct a higher-quality repository-level dataset comprising 6,816 CVSS labeled instances spanning 90 Common Weakness Enumeration (CWE) types. Experiments on real-world vulnerabilities show that CoSA consistently outperforms function-level and pure-LLM baselines. It improves prediction accuracy by 14.4% and Macro-F1 by 15.3% over the best-performing baseline, suggesting that explicit, metric-oriented repository context retrieval is crucial for practical and reliable automated severity assessment.

cs.CR↗

Minimaxity and Efficiency in Exponential Family Regression: From Star-Shaped to Convex Constraints

This paper establishes the minimax estimation rate for nonparametric exponential family regression under star-shaped constraints. We consider a parameter space $K$ that is a star-shaped subset of the hypercube $[-M, M]^n$ for a known constant $M > 0$. We operate under the assumption that the underlying exponential family is nonsingular with a twice continuously differentiable log-partition function. Our main result demonstrates that the minimax rate of the $\ell _{2}$ error of such estimation problem is $ε^{*2} \wedge \operatorname{diam}(K)^2$ up to constants exclusively depending on $M$. Here, the critical radius $ε^*$ is defined as \begin{equation*} ε^* = \sup \{ε\left\lvert\right. ε^2 κ(M) \le \log N^{\text{loc}}(ε,c)\}, \end{equation*} where $N^{\text{loc}}(ε,c)$ denotes the local metric entropy of $K$, and $κ(M) > 0, c>0$ are constants depending only on $M$. Such minimax rate is established by a match between an information-theoretic lower bound and an upper bound implied by a theoretical algorithm. Furthermore, we investigate the computational aspects of this estimation problem. Under mildly stronger assumptions on the constraint set $K$, we propose a computationally efficient, polynomial-time algorithm. We prove that the resulting estimator achieves the minimax optimal rate up to poly-logarithmic factors in the dimension $n$ and the geometric parameters of $K$. Finally, to illustrate the efficacy of our framework, we derive the minimax optimal rates for some concrete examples.

math.ST↗

Graph Is the Verifier: Agentic Reinforcement Learning for Interprocedural Vulnerability Detection

Real-world vulnerabilities often span multiple functions, yet most learning-based detectors classify each function in isolation: on a sample of real CVEs, we find that 71.7% of vulnerable functions require evidence from outside the function to be classified correctly. Agentic reinforcement learning (RL) could close this gap by enabling a model to gather that evidence itself, but it lacks a reliable reward, since a reward defined on the final verdict alone can be obtained without performing any investigation. We propose VulAgentRL, an agentic RL framework for interprocedural vulnerability detection built on a Code Property Graph (CPG). The CPG serves two roles: at inference time the policy queries it for callers, callees, dataflow, and other queries, and at training time the same graph verifies the evidence the policy cites. Because every CPG node carries a persistent integer identifier, this verification is an exact comparison rather than a textual match, so the reward credits verdicts that are supported by evidence. We further initialize the policy by distilling teacher investigations, and show that this warm start is necessary, since RL cannot acquire tool-use behavior it never samples. Under a repository-level split that prevents leakage, VulAgentRL outperforms state-of-the-art baselines, including frontier models, on the strict pair-wise-correct metric while issuing fewer tool calls, and its advantage persists on an out-of-distribution corpus and under class imbalance.

cs.CR↗

Astrophysical Population Coordinates for Supermassive Black Hole Binaries in Pulsar Timing Array Inference

Pulsar timing arrays can probe the population physics of supermassive black-hole binaries through the nanohertz gravitational-wave background. We construct a phenomenological forward model that follows source abundance, binary residence time, the high-mass population, finite-source strain moments, and the pulsar timing response. The simulated observables constrain three standardized population coordinates: $β$, which controls the residence-time and spectral response; $ϕ_{eff}$, which describes source normalization after accounting for its covariance with $β$; and $m_{eff}$, which is dominated by the high-mass cutoff. In the evaluation ensemble, the posterior-mean correlations with the simulated values are $0.928$, $0.926$, and $0.884$, with central 90 per cent coverages of $0.938\pm0.015$, $0.871\pm0.021$, and $0.898\pm0.019$, respectively. Frequency-resolved observables are most important for $β$, and strain moments beyond a common-process power law provide sensitivity to the normalization and high-mass coordinates; the fourth strain moment identifies $m_{eff}$ with rare, massive binaries. These coordinates quantify the relative sensitivity of the adopted PTA summaries within this population model, for which nearby population realizations retain substantial posterior overlap.

astro-ph.HE↗

SecureVibeBench: Benchmarking Secure Vibe Coding of AI Agents via Reconstructing Vulnerability-Introducing Scenarios

Large language model-powered code agents are rapidly transforming software engineering, yet the security risks of their generated code have become a critical concern. Existing benchmarks have provided valuable insights, but they fail to capture scenarios in which vulnerabilities are actually introduced by human developers, making fair comparisons between humans and agents infeasible. We therefore introduce SecureVibeBench, a benchmark of 105 C/C++ secure coding tasks sourced from 41 projects in OSS-Fuzz for code agents. SecureVibeBench has the following features: (i) realistic task settings that require multi-file edits in large repositories, (ii)~aligned contexts based on real-world open-source vulnerabilities with precisely identified vulnerability introduction points, and (iii) comprehensive evaluation that combines functionality testing and security checking with both static and dynamic oracles. We evaluate 5 popular code agents like OpenHands, supported by 5 LLMs (e.g., Claude sonnet 4.5) on SecureVibeBench. Results show that current agents struggle to produce both correct and secure code, as even the best-performing one, produces merely 23.8\% correct and secure solutions on SecureVibeBench. Our code and data are on https://github.com/iCSawyer/SecureVibeBench.

cs.SE↗

Efficient Robust Constrained Signal Detection via Kolmogorov Width Approximations

Robust statistical inference often faces a severe computational-statistical gap when dealing with complex parameter spaces. We investigate minimax signal detection in the Gaussian sequence model under strong $ε$-contamination, where the signal belongs to a general prior constraint $K$. Existing optimal tests require computing the exact Kolmogorov $k$-width of $K$, a computationally intractable task for general non-trivial sets. We bridge this gap by proposing a polynomial-time testing framework that universally applies to balanced, type-2, and exactly 2-convex constraints. By leveraging a semidefinite programming relaxation and a modified ellipsoid method equipped with an approximate subgradient oracle, we efficiently approximate the Kolmogorov widths. Remarkably, our unconditional efficient algorithm achieves a robust detection boundary that matches existing upper bounds up to a mere polylogarithmic factor. This establishes a computationally tractable testing solution for a broad class of structured signals without requiring prior knowledge of their exact geometric complexity.

math.ST↗

An Execution-Verified Multi-Language Benchmark for Code Semantic Reasoning

Evaluating whether large language models (LLMs) can recover execution-relevant program structure, rather than only produce code that passes tests, remains an open problem. Existing code benchmarks emphasize test-passing outputs, from standalone programming tasks (HumanEval, MBPP, LiveCodeBench) to repository repair (SWE-Bench); this is useful, but offers limited diagnostic signal about which program semantics a model can recover from source. We introduce TraceEval, to our knowledge the first execution-verified, multi-language benchmark for code semantic reasoning: recovering a program's runtime call structure from source code. Unlike prior call-graph benchmarks that rely on static-tool output or hand-annotated ground truth, every positive edge in TraceEval is mechanically witnessed by validation execution, eliminating annotator disagreement and label noise for observed behavior. TraceEval consists of (i) 10,583 real-world programs (2,129 test, 8,454 train) extracted from 1,600+ open-source repositories across Python, JavaScript, and Java via an LLM-assisted harness-generation pipeline with tracer validation; and (ii) a reproducible pipeline that converts any open-source repository into new verified benchmark instances. We evaluate 10 LLMs at zero-shot on the held-out test split. The strongest model, Claude-Opus-4.6, reaches an average F1 of 72.9% across the three languages. To demonstrate the train split's utility as a supervision substrate, we fine-tune the Qwen2.5-Coder family on it: lifts of up to +55.6 F1 bring tuned Qwen2.5-Coder-32B to 71.2%, within 1.7 F1 of zero-shot Claude-Opus-4.6. We release the benchmark, pipeline, baselines, and a datasheet at https://github.com/yikun-li/TraceEva

cs.SE↗

Text Tells the Cost: Predicting and Analyzing Repayment Effort of Self-Admitted Technical Debt

Technical debt refers to the consequences of sub-optimal decisions made during software development that prioritize short-term benefits over long-term maintainability. Self-Admitted Technical Debt (SATD) is a specific form of technical debt, explicitly documented by developers within software artifacts such as source code comments and commit messages. As SATD can hinder software development and maintenance, it is crucial to estimate the effort required to repay it so that we can effectively prioritize it. However, we currently lack an understanding of SATD repayment, and more importantly, we lack approaches that can automatically estimate the repayment effort of SATD based on its textual description. To bridge this gap, we have curated a comprehensive dataset of 341,740 SATD items from 2,568,728 commits across 1,060 Apache repositories and analyzed the repayment effort comparing SATD vs. non-SATD items, as well as different types of SATD items. Furthermore, we proposed an innovative approach for Predicting Repayment Effort of SATD using Textual Information, named PRESTI. Our findings show that different types of SATD require varying levels of repayment effort, with code/design, requirement, and test debt demanding greater effort compared to non-SATD items, while documentation debt requires less. We have evaluated our approaches, particularly BERT- and TextCNN-based models, which outperform traditional machine learning methods and the baseline in estimating repayment effort. Additionally, we summarize keywords associated with varying levels of repayment effort that occur during SATD repayment. Our work aims to enhance SATD repayment prioritization and resource allocation, thereby improving software development and maintainability.

cs.SE↗

Semantics-Aligned, Curriculum-Driven, and Reasoning-Enhanced Vulnerability Repair Framework

Current learning-based Automated Vulnerability Repair (AVR) approaches, while promising, often fail to generalize effectively in real-world scenarios. Our diagnostic analysis reveals three fundamental weaknesses in state-of-the-art AVR approaches: (1) limited cross-repository generalization, with performance drops on unseen codebases; (2) inability to capture long-range dependencies, causing a performance degradation on complex, multi-hunk repairs; and (3) over-reliance on superficial lexical patterns, leading to significant performance drops on vulnerabilities with minor syntactic variations like variable renaming. To address these limitations, we propose SeCuRepair, a semantics-aligned, curriculum-driven, and reasoning-enhanced framework for vulnerability repair. At its core, SeCuRepair adopts a reason-then-edit paradigm, requiring the model to articulate why and how a vulnerability should be fixed before generating the patch. This explicit reasoning enforces a genuine understanding of repair logic rather than superficial memorization of lexical patterns. SeCuRepair also moves beyond traditional supervised fine-tuning and employs semantics-aware reinforcement learning, rewarding patches for their syntactic and semantic alignment with the oracle patch rather than mere token overlap. Complementing this, a difficulty-aware curriculum progressively trains the model, starting with simple fixes and advancing to complex, multi-hunk coordinated edits. We evaluate SeCuRepair on strict, repository-level splits of BigVul and newly crafted PrimeVul_AVR datasets. SeCuRepair significantly outperforms all baselines, surpassing the best-performing baselines by 34.52% on BigVul and 31.52% on PrimeVul\textsubscript{AVR} in terms of CodeBLEU, respectively. Comprehensive ablation studies further confirm that each component of our framework contributes to its final performance.

cs.SE↗

Revisiting Vulnerability Patch Identification on Data in the Wild

Attacks can exploit zero-day or one-day vulnerabilities that are not publicly disclosed. To detect these vulnerabilities, security researchers monitor development activities in open-source repositories to identify unreported security patches. The sheer volume of commits makes this task infeasible to accomplish manually. Consequently, security patch detectors commonly trained and evaluated on security patches linked from vulnerability reports in the National Vulnerability Database (NVD). In this study, we assess the effectiveness of these detectors when applied in-the-wild. Our results show that models trained on NVD-derived data show substantially decreased performance, with decreases in F1-score of up to 90\% when tested on in-the-wild security patches, rendering them impractical for real-world use. An analysis comparing security patches identified in-the-wild and commits linked from NVD reveals that they can be easily distinguished from each other. Security patches associated with NVD have different distribution of commit messages, vulnerability types, and composition of changes. These differences suggest that NVD may be unsuitable as the \textit{sole} source of data for training models to detect security patches. We find that constructing a dataset that combines security patches from NVD data with a small subset of manually identified security patches can improve model robustness.

cs.SE↗

Robust Signal Detection with Quadratically Convex Orthosymmetric Constraints

This paper studies the problem of robust signal detection in Gaussian noise under quadratically convex orthosymmetric (QCO) constraints. We consider a minimax testing framework where the signal belongs to a QCO set and is separated from zero in Euclidean norm, while an adversary is allowed to arbitrarily corrupt a fraction $ε$ of the samples. We establish the minimax separation radius between the null and alternative purely in terms of the constraint geometry, sample size, corruption rate, and noise scale. Our analysis argues that the Kolmogorov widths of the constraint set play a central role in determining the detection limits, paralleling to classic results in estimation problem. The derived lower bounds exhibit phase transitions with respect to the corruption rate and confirm that robust testing is statistically easier than robust estimation. While the information-theoretic upper bound is achieved by a computationally intractable test, we develop a polynomial-time algorithm that achieves the minimax lower bound up to logarithmic factors. Unlike prior work, our algorithm handles signals of arbitrary Euclidean length while respecting the QCO constraints. Finally, we extend these results to the robust $\ell _{p}$ norm testing for $1 \le p < 2$.

math.ST↗

Real time filtering algorithms

This paper presents a systematic review of recent advances in nonlinear filtering algorithms, structured into three principal categories: Kalman-type methods, Monte Carlo methods, and the Yau-Yau algorithm. For each category, we provide a comprehensive synthesis of theoretical developments, algorithmic variants, and practical applications that have emerged in recent years. Importantly, this review addresses both continuous-time and discrete-time system formulations, offering a unified review of filtering methodologies across different frameworks. Furthermore, our analysis reveals the transformative influence of artificial intelligence breakthroughs on the entire nonlinear filtering field, particularly in areas such as learning-based filters, neural network-augmented algorithms, and data-driven approaches.

math.OC↗

Beyond Function-Level Analysis: Context-Aware Reasoning for Inter-Procedural Vulnerability Detection

Recent progress in ML and LLMs has improved vulnerability detection, and recent datasets have reduced label noise and unrelated code changes. However, most existing approaches still operate at the function level, where models are asked to predict whether a single function is vulnerable without inter-procedural context. In practice, vulnerability presence and root cause often depend on contextual information. Naively appending such context is not a reliable solution: real-world context is long, redundant, and noisy, and we find that unstructured context frequently degrades the performance of strong fine-tuned code models. We present CPRVul, a context-aware vulnerability detection framework that couples Context Profiling and Selection with Structured Reasoning. CPRVul constructs a code property graph, and extracts candidate context. It then uses an LLM to generate security-focused profiles and assign relevance scores, selecting only high-impact contextual elements that fit within the model's context window. In the second phase, CPRVul integrates the target function, the selected context, and auxiliary vulnerability metadata to generate reasoning traces, which are used to fine-tune LLMs for reasoning-based vulnerability detection. We evaluate CPRVul on three high-quality vulnerability datasets: PrimeVul, TitanVul, and CleanVul. Across all datasets, CPRVul consistently outperforms function-only baselines, achieving accuracies ranging from 64.94% to 73.76%, compared to 56.65% to 63.68% for UniXcoder. Specifically, on the challenging PrimeVul benchmark, CPRVul achieves 67.78% accuracy, outperforming prior state-of-the-art approaches, improving accuracy from 55.17% to 67.78% (22.9% improvement). Our ablations further show that neither raw context nor processed context alone benefits strong code models; gains emerge only when processed context is paired with structured reasoning.

cs.CR↗