arXiv subjects

Danqing Wang

Publications and source records attributed to Danqing Wang.

At least 19 recordsLinked to original sources

arXiv2026

Ulysses: A Security Harness For Synthesizing Verifiable Web Exploits

How can an AI agent automatically discover security vulnerabilities in any software repository? Existing approaches use LLMs to generate software vulnerability exploits. However, it is challenging for an LLM to reliably verify its own generated exploits. In this paper, we propose Ulysses, a multi-agent security harness to synthesize verifiable vulnerability exploits for web applications. Ulysses converts candidate vulnerability findings into verified Proofs of Exploitability by iteratively exploring exploit hypotheses and establishing success in independently observed execution effects. We evaluated Ulysses on WebExploitGym, a curated benchmark of exploitable instances from 25 real-world web applications that span 5 languages and 7 frameworks. Ulysses produces verified proofs for 10 of 25 real-world advisories. It also discovered 4 new proofs of exploitability that have never been previously reported.

cs.CR↗

arXiv2026

Self-Evolving Coding Agents via Harness-Policy Optimization

Large language model (LLM) agents have shown promising potential in coding tasks, and their performance is jointly determined by the policy model and the harness that mediates their interaction with the environment. Indeed, harness and policy are mutually reinforcing and evolving only one component is suboptimal---a harness tuned for one checkpoint may not target the failure modes of an updated policy, while policy learning remains limited to the trajectories elicited by a fixed harness. In light of this, we propose Self-HPO, a recursive self-improving framework for coding agents that alternates harness and policy optimization, using the evolving policy as its own harness proposer. Specifically, in each round, the current policy first analyzes its own failures and proposes new harness candidates (H-step). The policy is then updated through reinforcement learning (RL) under the selected harness (P-step), and hard tasks identified during training form the next harness search set, keeping harness optimization targeted to the policy's evolving failure modes. Across three coding benchmarks, Self-HPO consistently outperforms harness-only and policy-only baselines, improving Qwen3.5-35B-A3B's average resolve rate from 57.8% to 64.8%, with gains generalizing to out-of-distribution settings. Further analysis shows that both optimized components retain gains even when paired with their base counterparts, while their joint gain exceeds the sum of their individual gains. Self-HPO also improves reliability (Pass@1) and expands the capability frontier (Pass@k) over the base agent. Additionally, we show that with a stronger external model (e.g., Opus 4.8) as the harness proposer, the same framework yields a further performance boost.

cs.CL↗

arXiv2026

AuraForge: Scaling Security Supervision for Training Coding Agents

Coding agents are now proficient enough to generate complex software applications from a single prompt. As their capabilities have grown, human oversight has increasingly shifted from line-by-line code review toward hands-off evaluation of outcomes. However, recent studies have shown that such a transition exposes a critical risk: functional correctness alone does not guarantee a secure implementation. Despite growing attention to code security, training safer coding agents remains challenging because reliable security supervision is difficult to obtain at scale from real-world repositories. We introduce AuraForge to synthesize and validate executable security tests for training secure coding agents. Our approach combines attack-oriented test synthesis, language-extensible task construction, and safeguards against reward hacking. Using AuraForge, we construct AuraGym, a multi-language and multi-CWE executable training gym: 679 executable feature-implementation tasks from 344 real-world repositories across Python, JavaScript, and TypeScript, covering 177 CWE categories. On the subset with human-written security tests, AuraForge produces about 3 times as many test cases on average and reduces the false-positive rate by 83.23%, allowing alternative secure implementations to receive correct supervision. Training Qwen3.5-4B with synthesized security tests gains larger improvements than human-written security tests (average 19.7 FuncPass and 6.2 SecPass vs. 14.9 FuncPass and 4.4 SecPass) on three languages. These results demonstrate that AuraForge provides more diverse and reliable security supervision to train secure coding agents.

cs.CR↗

arXiv2026

ABSENTIA: Detecting Broken Access Control Vulnerabilities in Web Applications

Broken access control, the failure of authorization, is one of the most prevalent web security risks. Unlike injection, a flow of untrusted input into a dangerous operation, authorization is a relation: who may act on what, not how data moves. Each application decides that relation for itself, so no rule written in advance carries to the next. An LLM agent can infer it from the code, but with no systematic way to cover the application and prioritize what to inspect, its search stays undirected and access-control flaws go undetected. We present ABSENTIA, a security scaffolding that turns general LLM agents into systematic vulnerability detectors for the backend of web applications, run as an audit by the developers and security engineers who maintain the code. Under its direction, the agents build a graph that maps the application's routes to the code behind them. ABSENTIA then works route by route, applying invariant falsification: it infers the properties the code is meant to satisfy, and where one is not enforced, reports the route for maintainer review. We also release BAC-Bench, a benchmark of 30 broken access control advisories across 25 repositories, 3 languages, and 9 frameworks, each published in 2025 or later, verified by a human auditor, and paired with its fixing commit, so credit requires flagging the vulnerable version and not the fixed one. ABSENTIA recalls 19 of them, 17 under paired credit, and an LLM verifier confirms 51% of its findings. CodeQL and Semgrep recall none, and an unstructured agent on the same model recalls 3. In the OWASP Benchmark injection categories, ABSENTIA leads the dedicated analyzers in Python and trails only CodeQL and IRIS in Java.

cs.CR↗

arXiv2026

SecureVibe: Making Vibe Coding More Secure

As vibe coding becomes increasingly capable and widespread, security vulnerabilities in even functionally correct solutions are a growing concern. When investigating functionally correct but insecure solutions, we find that the insecure agent is less than half as likely to conduct effective planning and testing for the hidden security risks behind the functional requirements. Motivated by this, we develop SECUREVIBE, a training recipe that explicitly targets planning and testing for code security. SECUREVIBE constructs training signals around these security behaviors. It includes supervised fine-tuning on the security suite with 4 security tasks, and post-training methods, SECUREVIBE_rl and SECUREVIBE_hg, to enhance security capabilities from verifiable execution feedback and hint-based self-supervision. Our SECUREVIBE outperforms the baseline on two types of security coding tasks across 4 benchmarks. Specifically, SECUREVIBE improves the security pass@1 by 6.9 points on BaxBench. The gains extend to unseen CWE categories, with improvements of 11.5 points on SusVibes. Meanwhile, it also improves functionality pass@1 by 13.6 points on the security coding task SusVibes and 4.1 points on the generic coding task SWE-bench Verified. Further analysis offers two practical insights: (i) diversifying supervision across security planning, coding, and testing strengthens security behaviors more effectively than adding coding trajectories alone, and (ii) hint-guided supervision is particularly valuable when the agent's existing security capabilities are insufficient to learn effectively from outcome feedback.

cs.CR↗

arXiv2026

Is Vibe Coding Safe? Benchmarking Vulnerability of Agent-Generated Code in Real-World Tasks

Vibe coding is a new software development paradigm in which human engineers prompt a large language model (LLM) agent to complete complex coding tasks with little supervision. Although vibe coding is increasingly adopted, is the generated code really safe to deploy in production? To investigate this question, we propose SUSVIBES, a benchmark consisting of 186 feature-request software engineering tasks from real-world open-source projects, for which, human programmers committed vulnerable implementations. We evaluate 12 widely used coding agentic settings with frontier models on the benchmark. Disturbingly, all agents perform poorly in terms of software security. Although 57% of the solutions from SWE-Agent with Claude 4 Sonnet are functionally correct, only 11.8% are secure. Further experiments demonstrate that preliminary security strategies, such as augmenting the feature request with vulnerability hints, cannot mitigate these security issues. Our findings raise serious concerns about the widespread adoption of vibe coding, particularly in security-sensitive applications. The code and dataset are available at https://github.com/LeiLiLab/susvibes. The leaderboard is at https://leililab.github.io/susvibes-leaderboard.

cs.SE↗

arXiv2026

Learning to Interrupt in Language-based Multi-agent Communication

When a colleague starts explaining something you already understand, you interrupt them. This simple act, a listener taking control of the conversation, is natural in human communication but absent in current verbose LLM multi-agent systems. Current approaches address communication efficiency only from the speaker side, compressing messages before they are sent. We flip the perspective: rather than making speakers more concise, we let listeners decide when they have heard enough. We propose a new communication paradigm in which the listener can interrupt the speaker mid-generation. We find that LLMs, given this ability, are overconfident and interrupt too early before receiving sufficient information. This finding motivates HANDRAISER, a learning method that predicts the right moment to interrupt based on estimated future reward and communication cost. We evaluate our framework on three multi-agent tasks: 2-agent text pictionary, 3-agent meeting scheduling, and 3-agent debate. HANDRAISER reduces communication cost by 32.2% over the non-interruptible baseline while achieving comparable or superior task performance, with interruption behavior that generalizes across different agents and tasks.

cs.CL↗

arXiv2026

A K-band Kinetic Inductance Parametric Amplifier Near the Quantum Limit

Advancing superconducting quantum devices to higher operating frequencies broadens their functionality and enables operation at elevated temperatures, but it also requires near-quantum-limited amplifiers beyond the few-gigahertz regime. Here we present a junction-free, kinetic-inductance parametric amplifier based on thin-film niobium nitride (NbN) operating at 23 GHz in the microwave K-band, achieving a gain up to 40 dB, a 100 MHz gain-bandwidth product, a 1 dB saturation input power of -85 dBm with 23 dB gain, and added noise no greater than 1.4 quanta for phase-preserving amplification. Leveraging the large superconducting gap of NbN, this architecture can be extended to even higher frequencies, supporting applications such as high-fidelity readout of millimeter-wave superconducting qubits and axion searches over an expanded mass window.

quant-ph↗

arXiv2026

CRAB-Bench: Evaluating LLM Agents under Complex Task Dependencies and Human-aligned User Simulation

Evaluating LLM agents in realistic service scenarios requires complex task dependencies, imperfect user behavior, and an evaluation that accommodates multiple valid solutions. We introduce CRAB-Bench (Constraint-based Realistic Agent Benchmark) and RUSE (Realistic User Simulation Engine) to address this gap. CRAB-Bench generates tasks via a constraint graph over multiple interdependent entities with structured distractors, requiring agents to reason carefully over thousands of misleading candidates where only a tiny fraction of solutions are valid. RUSE replaces cooperative, template-like simulators with realistic users grounded in human behavioral studies, instantiated across diverse personas and four behavioral dimensions. Experiments on four frontier LLM agents show that the best model achieves only 61% pass@1 on CRAB-Bench, and switching to RUSE causes further drops of up to 57%, concentrated in task-solving ability rather than conversational quality. Information Disclosure is the most damaging behavioral dimension, and agents interacting with RUSE are less likely to admit mistakes, instead masking errors through implicit corrections.

cs.CL↗

arXiv2026

A frequency-agile microwave-optical interface for superconducting qubits

Superconducting quantum processors operate at microwave frequencies in millikelvin environments, making it challenging to interconnect distant nodes using conventional microwave wiring. Coherent microwave-to-optical (M2O) transduction enables superconducting quantum networks by interfacing itinerant microwave photons with low-loss optical fiber. However, many state-of-the-art transducers provide efficient conversion only over a narrow frequency span, complicating deployment with heterogeneous superconducting devices that are detuned by gigahertz-scale offsets. Here we demonstrate a frequency-agile microwave-optical interface that overcomes this bandwidth mismatch by cascading an electro-optic M2O transducer with a multimode microwave-to-microwave (M2M) frequency converter, with in situ tunability of the microwave resonances in both stages. Using this architecture, we realize continuous frequency coverage from 5.0 to 8.5 GHz within a single system. As an application relevant to superconducting-qubit networking, we use the cascaded M2M-M2O interface to optically read out a superconducting qubit whose readout resonator is detuned by 1.7 GHz from the native M2O microwave resonance, demonstrating a scalable route toward fiber-linked superconducting quantum nodes.

quant-ph↗

arXiv2025

Strategic Planning and Rationalizing on Trees Make LLMs Better Debaters

Winning competitive debates requires sophisticated reasoning and argument skills. There are unique challenges in the competitive debate: (1) The time constraints force debaters to make strategic choices about which points to pursue rather than covering all possible arguments; (2) The persuasiveness of the debate relies on the back-and-forth interaction between arguments, which a single final game status cannot evaluate. To address these challenges, we propose TreeDebater, a novel debate framework that excels in competitive debate. We introduce two tree structures: the Rehearsal Tree and Debate Flow Tree. The Rehearsal Tree anticipates the attack and defenses to evaluate the strength of the claim, while the Debate Flow Tree tracks the debate status to identify the active actions. TreeDebater allocates its time budget among candidate actions and uses the speech time controller and feedback from the simulated audience to revise its statement. The human evaluation on both the stage-level and the debate-level comparison shows that our TreeDebater outperforms the state-of-the-art multi-agent debate system, with a +15.6% improvement in stage-level persuasiveness with DeepSeek and +10% debate-level opinion shift win. Further investigation shows that TreeDebater shows better strategies in limiting time to important debate actions, aligning with the strategies of human debate experts.

cs.CL↗

arXiv2025

Hidden Structural Variants in ALD NbN Superconducting Trilayers Revealed by Atomistic Analysis

Microscopic inhomogeneity within superconducting films is a critical bottleneck hindering the performance and scalability of quantum circuits. All-nitride Josephson Junctions (JJs) have attracted substantial attention for their potential to provide enhanced coherence times and enable higher temperature operation. However, their performance is often limited by local variations caused by polymorphism, impurities, and interface quality. This work diagnoses atomic-scale limitations preventing superconducting NbN/AlN/NbN JJs from reaching their full potential. Electrical measurements reveal suppressed critical current density and soft onset of quasiparticle current. However, inverse proportionality between resistance and junction area confirms homogenous barrier thickness. This isolates structural and chemical variations in electrodes and barrier as the source of performance limitation. The observed characteristics are attributed to complex materials problems: NbN polymorphism, phase coexistence, and oxygen impurities. Using advanced microscopy and machine learning integrated approach, nanoscale inclusions of epsilon-Nb2N2 are found to coexist within dominant delta-NbN electrodes. DC performance of JJs may be affected by these defects, leading to unresolved supercurrent and soft transition to normal state. By identifying specific atomic scale defects, tracing its origin to initial film nucleation, and linking to its detrimental electrical signature, this work establishes a material-to-device correlation and provides targeted strategy for phase engineering towards reproducible, high coherence and scalable quantum devices.

quant-ph↗

arXiv2025

All-nitride superconducting qubits based on atomic layer deposition

The development of large-scale quantum processors benefits from superconducting qubits that can operate at elevated temperatures and be fabricated with scalable, foundry-compatible processes. Atomic layer deposition (ALD) is increasingly being adopted as an industrial standard for thin-film growth, particularly in applications requiring precise control over layer thickness and composition. Here, we report superconducting qubits based on NbN/AlN/NbN trilayers deposited entirely by ALD. By varying the number of ALD cycles used to form the AlN barrier, we achieve Josephson tunneling through barriers of different thicknesses, with critical current density spanning seven orders of magnitude, demonstrating the uniformity and versatility of the process. Owing to the high critical temperature of NbN, transmon qubits based on these all-nitride trilayers exhibit microsecond-scale relaxation times, even at temperatures above 300 mK. These results establish ALD as a viable low-temperature deposition technique for superconducting quantum circuits and position all-nitride ALD qubits as a promising platform for operation at elevated temperatures.

quant-ph↗

arXiv2025

RARE: Retrieval-Aware Robustness Evaluation for Retrieval-Augmented Generation Systems

Retrieval-Augmented Generation (RAG) enhances recency and factuality in answers. However, existing evaluations rarely test how well these systems cope with real-world noise, conflicting between internal and external retrieved contexts, or fast-changing facts. We introduce Retrieval-Aware Robustness Evaluation (RARE), a unified framework and large-scale benchmark that jointly stress-tests query and document perturbations over dynamic, time-sensitive corpora. One of the central features of RARE is a knowledge-graph-driven synthesis pipeline (RARE-Get) that automatically extracts single and multi-hop relations from the customized corpus and generates multi-level question sets without manual intervention. Leveraging this pipeline, we construct a dataset (RARE-Set) spanning 527 expert-level time-sensitive finance, economics, and policy documents and 48295 questions whose distribution evolves as the underlying sources change. To quantify resilience, we formalize retrieval-conditioned robustness metrics (RARE-Met) that capture a model's ability to remain correct or recover when queries, documents, or real-world retrieval results are systematically altered. Our findings reveal that RAG systems are unexpectedly sensitive to perturbations. Moreover, they consistently demonstrate lower robustness on multi-hop queries compared to single-hop queries across all domains.

cs.CL↗

arXiv2025

TypedThinker: Diversify Large Language Model Reasoning with Typed Thinking

Large Language Models (LLMs) have demonstrated strong reasoning capabilities in solving complex problems. However, current approaches primarily enhance reasoning through the elaboration of thoughts while neglecting the diversity of reasoning types. LLMs typically employ deductive reasoning, proceeding step-by-step from given conditions, which limits their exploration during problem-solving. Our analysis reveals that certain problems are exclusively solvable through specific reasoning strategies like inductive, abductive, or analogical reasoning. However, incorporating diverse reasoning approaches presents two key challenges: identifying the appropriate reasoning type for each problem and exploiting this approach during problem-solving. Therefore, we propose the TypedThinker that predicts suitable reasoning types based on the problem and their previous effectiveness and provides relevant demonstrations to guide LLMs in applying these strategies. Experimental results show significant improvements across multiple benchmarks, with performance gains of 3.4% for Mistral 7B, 6.5% for LLaMA3 8B, and 7% for Qwen 2 7B on logical and mathematical reasoning tasks. TypedThinker enhances LLM reasoning without requiring knowledge distillation from larger models. It can be integrated into more advanced systems like GPT-4o or specialized models like MetaMath to diversify their reasoning approaches and improve their problem-solving capabilities.

cs.CL↗

arXiv2025

Focus Directions Make Your Language Models Pay More Attention to Relevant Contexts

Long-context large language models (LLMs) are prone to be distracted by irrelevant contexts. The reason for distraction remains poorly understood. In this paper, we first identify the contextual heads, a special group of attention heads that control the overall attention of the LLM. Then, we demonstrate that distraction arises when contextual heads fail to allocate sufficient attention to relevant contexts and can be mitigated by increasing attention to these contexts. We further identify focus directions, located at the key and query activations of these heads, which enable them to allocate more attention to relevant contexts without explicitly specifying which context is relevant. We comprehensively evaluate the effect of focus direction on various long-context tasks and find out focus directions could help to mitigate the poor task alignment of the long-context LLMs. We believe our findings could promote further research on long-context LLM alignment.

cs.CL↗

arXiv2025

Broad Spectrum Coherent Frequency Conversion with Kinetic Inductance Superconducting Metastructures

Parametric frequency converters (PFCs) play a critical role in bridging the frequency gap between quantum information carriers. PFCs in the microwave band are particularly important for superconducting quantum processors, but their operating bandwidth is often strongly limited. Here, we present a multimode kinetic metastructure for parametric frequency conversion between broadly spanning frequency modes. This device comprises a chain of asymmetric kinetic inductance grids designed to deliver efficient three-wave mixing nonlinearity. We demonstrate high efficient coherent conversion among broadly distributed modes, and the mode frequency is continuously tunable by controlling the external magnetic field strength, making it ideally suited for quantum computing and communication applications requiring flexible and efficient frequency conversion.

quant-ph↗

arXiv2024

Scaling LLM Inference with Optimized Sample Compute Allocation

Sampling is a basic operation in many inference-time algorithms of large language models (LLMs). To scale up inference efficiently with a limited compute, it is crucial to find an optimal allocation for sample compute budgets: Which sampling configurations (model, temperature, language, etc.) do we use? How many samples do we generate in each configuration? We formulate these choices as a learning problem and propose OSCA, an algorithm that Optimizes Sample Compute Allocation by finding an optimal mix of different inference configurations. Our experiments show that with our learned mixed allocation, we can achieve accuracy better than the best single configuration with 128x less compute on code generation and 25x less compute on 4 reasoning tasks. OSCA is also shown to be effective in agentic workflows beyond single-turn tasks, achieving a better accuracy on SWE-Bench with 3x less compute than the default configuration. Our code and generations are released at https://github.com/LeiLiLab/OSCA.

cs.CL↗