Searcharxiv⌕ Search

arXiv subjects

Liang Cheng

Publications and source records attributed to Liang Cheng.

At least 19 recordsLinked to original sources
arXiv2026

ASKAP discovery of a pair of large radio bubbles: on the origin of odd radio circles

We report the serendipitous discovery of a large, low-surface-brightness radio bubble in 944 MHz continuum data from the ASKAP Evolutionary Map of the Universe (EMU) survey. The structure, centred on the elliptical galaxy LEDA 217397 at a redshift of $z=0.040$, spans $\sim$8.4 arcmin, corresponding to a projected diameter of $\sim$399 kpc, and consists of two partly overlapping shells both with radii of $\sim$114 kpc. The integrated flux density of the bubble is $56.8\pm2.9$ mJy at 944 MHz, implying a rest-frame 1.4 GHz luminosity of $\sim$ $1.4\times10^{23}$ W Hz$^{-1}$. Combining the EMU measurement with MWA GLEAM-X data at 88--185 MHz, we derive a steep integrated spectral index of $\alpha=-1.04\pm0.04$, and a two-frequency spectral-index map suggesting a possible exterior flattening. Spectral Energy Distribution (SED) fitting indicates a massive ($\log M_{\ast}/M\odot = 10.97\pm0.09$), quiescent (SFR=$0.025\pm0.083\,M\odot$ yr$^{-1}$) early-type host with no mid-infrared AGN signature and no overdense environment. We compare the bubble with odd radio circles (ORCs) and large radio shells, and discuss three scenarios for its origin: a starburst-driven wind, a merger-driven shock, and AGN jet-inflated bubbles. The starburst wind is disfavoured on energetic grounds ($\gtrsim$$10^{59}$ erg required versus $\sim$$10^{8}$ yr electron lifetimes), and neither a halo-scale merger shock nor a spherical nuclear blast wave can explain the unusually regular, double-shell geometry; a bipolar nuclear outburst -- a relic AGN jet episode, possibly triggered by a supermassive-black-hole merger -- provides the most natural explanation, with later shocks possibly re-energising the plasma. Deeper broad-band radio, polarimetric, spectroscopic and X-ray observations are needed to confirm its nature.

astro-ph.GA↗
arXiv2026

Ball and Spherical-Shell Rigidity from Overdetermined Translating Solitons

We study overdetermined boundary problems for the graphical translating-soliton equation \[ -\operatorname{div}\!\left(\frac{Du}{\sqrt{1+|Du|^2}}\right) =\frac{1}{\sqrt{1+|Du|^2}} \quad\text{in }\Omega, \qquad \partial_\nu u=\Gamma H+C \quad\text{on }\partial\Omega, \] where $\Gamma, C$ are constants, $H$ is the mean curvature of the boundary $\partial\Omega$ of the regular bounded domain in $R^n$ such that $H_{\partial B_R}=-1/R$. For $\Gamma\geq0$, we prove that constant Dirichlet data force a bounded domain to be a ball. We also prove a spherical-shell rigidity theorem for a doubly connected domain with two ordered boundary heights and $a<u<b$ in the interior. The argument combines linearization under reflection, Reichel's critical-plane and annular continuation principles, curvature comparison, Serrin's corner lemma, and a radial ODE that excludes the annular alternative in the one-height problem. Finally, we give explicit counterexamples showing the sharpness of the sign, ordering, connectedness, and nesting assumptions.

math.AP↗
arXiv2026

Specialize Roles, Mix Deployments: Pushing the Cost-Accuracy Frontier of LLM Agent Teams

LLM agents are increasingly deployed as multi-role teams, where tasks are divided across specialized roles such as planner, executor, and verifier. In these systems, cost and accuracy are no longer properties of a single model: they depend on which model fills each role and where it is hosted, including API, self-hosted, and hybrid deployment. Existing agentic benchmarks typically evaluate fixed models or fixed agent configurations, and therefore offer limited guidance for cost-accuracy-optimal deployment. We introduce AgentCARD, a role-aware benchmark suite for evaluating LLM agent teams across role assignment and deployment mode. AgentCARD combines a role-decomposed evaluation harness, a unified API/self-hosted cost model, Pareto-frontier analysis, and a Shapley-based diagnostic for identifying role bottlenecks. Our evaluation shows that heterogeneous teams consistently occupy the cost-accuracy frontier. They improve accuracy by up to $44\%$ over cost-equivalent homogeneous teams, or match the strongest homogeneous team at up to $12\times$ lower per-task cost through hybrid deployment. We further find that the best role assignment is domain-dependent: some domains are planner-bottlenecked, while others are executor-bottlenecked. Finally, AgentCARD extends beyond planner--executor teams to workflows with additional roles such as verification, and supports continual evaluation as new domains and team structures emerge. Our code is released at: https://github.com/Auto-CAP/AgentCAP

cs.MA↗
arXiv2026

Do Agents Know What They Can't Do? Evaluating Feasibility Awareness in Tool-Using Agents

Tool-using agents often incur substantial computational cost due to long reasoning chains and iterative tool usage. In practical scenarios, many tasks become infeasible under constrained tool environments, where the capabilities required for successful task completion are unavailable. Detecting infeasible tasks and stopping execution early can significantly reduce unnecessary execution cost. In this work, we propose FeasiGen, an automatic pipeline for constructing infeasible agent tasks by identifying the critical tools required for successful task completion. Our approach extracts tool-calling traces from successful executions across multiple agent systems, identifies critical tools consistently shared across diverse execution strategies, and masks these tools to automatically transform solvable tasks into infeasible ones. Human verification confirms that the infeasibility annotations for our constructed tasks achieve over 94% accuracy. We further introduce feasibility-aware evaluation metrics for measuring whether agents can recognize infeasible tasks and stop execution appropriately. Extensive evaluations across nine models reveal substantially weak infeasibility detection ability, with false continue rate reaching up to 73.9%. We further observe that multi-agent architectures significantly reduce erroneous execution under infeasible conditions.

cs.AI↗
arXiv2025

ContextPilot: Fast Long-Context Inference via Context Reuse

AI applications increasingly depend on long-context inference, where LLMs consume substantial context to support stronger reasoning. Common examples include retrieval-augmented generation, agent memory layers, and multi-agent orchestration. As input contexts get longer, prefill latency becomes the main bottleneck. Yet today's prefill acceleration techniques face a trade-off: they either preserve reasoning quality but deliver little KV-cache reuse, or improve reuse at the cost of degraded reasoning quality. We present ContextPilot, a system that accelerates prefill by introducing context reuse as a new mechanism for faster long-context inference. ContextPilot introduces a context index to identify overlapping context blocks across LLM interactions (e.g., across users and turns). It further proposes context ordering and de-duplication techniques to maximize KV-cache reuse. To preserve reasoning quality under reuse, it introduces succinct context annotations that prevent quality degradation. Finally, ContextPilot is built around a modular architecture with a clean interface that integrates with existing inference engines. Extensive evaluation shows that ContextPilot reduces LLM prefill latency by up to $3\times{}$ compared to state-of-the-art methods while preserving reasoning quality. At longer context lengths, it can even improve reasoning quality. ContextPilot is open-sourced at: https://github.com/EfficientContext/ContextPilot.

cs.LG↗
arXiv2025

Psyzkaller: Learning from Historical and On-the-Fly Execution Data for Smarter Seed Generation in OS kernel Fuzzing

OS Kernel fuzzers such as Syzkaller often struggle to generate syscall sequences that respect intrinsic Syscall Dependency Relations (SDRs), resulting in seeds that either violate kernel constraints or fail to reach deep execution paths. We propose leveraging an N-gram model to learn SDRs from both kernel execution history and ongoing fuzzing results. This enables the fuzzer to capture dependencies in similar kernel versions while adapting to target-specific behaviors, thereby improving the validity of generated seeds. Additionally, we introduce a bidirectional Random Walk strategy to enhance the diversity of generated seeds. We implement this approach in a prototype, Psyzkaller, on top of Syzkaller. Experiments show that, trained with the large-scale DongTing dataset and continuously updated with ongoing fuzzing results, Psyzkaller improves Syzkaller's code coverage by 4.6%-7.0%, triggers 110.4%-187.2% more crashes, and discovers eight previously unknown kernel vulnerabilities. Furthermore, Psyzkaller outperforms state-of-the-art fuzzers such as ACTOR and SyzDescribe in both coverage and crashes.

cs.CR↗
arXiv2025

Automatic Generation of a Cryptography Misuse Taxonomy Using Large Language Models

The prevalence of cryptographic API misuse (CAM) is compromising the effectiveness of cryptography and in turn the security of modern systems and applications. Despite extensive efforts to develop CAM detection tools, these tools typically rely on a limited set of predefined rules from human-curated knowledge. This rigid, rule-based approach hinders adaptation to evolving CAM patterns in real practices. We propose leveraging large language models (LLMs), trained on publicly available cryptography-related data, to automatically detect and classify CAMs in real-world code to address this limitation. Our method enables the development and continuous expansion of a CAM taxonomy, supporting developers and detection tools in tracking and understanding emerging CAM patterns. Specifically, we develop an LLM-agnostic prompt engineering method to guide LLMs in detecting CAM instances from C/C++, Java, Python, and Go code, and then classifying them into a hierarchical taxonomy. Using a data set of 3,492 real-world software programs, we demonstrate the effectiveness of our approach with mainstream LLMs, including GPT, Llama, Gemini, and Claude. It also allows us to quantitatively measure and compare the performance of these LLMs in analyzing CAM in realistic code. Our evaluation produced a taxonomy with 279 base CAM categories, 36 of which are not addressed by existing taxonomies. To validate its practical value, we encode 11 newly identified CAM types into detection rules and integrate them into existing tools. Experiments show that such integration expands the tools' detection capabilities.

cs.CR↗
arXiv2025

A Survey of Graph Neural Networks for Drug Discovery: Recent Developments and Challenges

Graph Neural Networks (GNNs) have gained traction in the complex domain of drug discovery because of their ability to process graph-structured data such as drug molecule models. This approach has resulted in a myriad of methods and models in published literature across several categories of drug discovery research. This paper covers the research categories comprehensively with recent papers, namely molecular property prediction, including drug-target binding affinity prediction, drug-drug interaction study, microbiome interaction prediction, drug repositioning, retrosynthesis, and new drug design, and provides guidance for future work on GNNs for drug discovery.

cs.LG↗
arXiv2025

Statistical Invisibility of a Physical Attack on QRNGs After Randomness Extraction

Current prevailing designs of quantum random number generators (QRNGs) designs typically employ post-processing techniques to distill raw random data, followed by statistical verification with suites like NIST SP 800-22. This paper demonstrates that this widely adopted practice harbors a critical flaw. We show that the powerful extraction process can create a false sense of security by perfectly concealing physical-layer attacks, rendering the subsequent statistical tests blind to a compromised entropy source. We substantiate this claim across two major QRNG architectures. Experimentally, we severely compromise an QRNG based on amplified spontaneous emission (ASE) with a power supply ripple attack. While the resulting raw data catastrophically fails NIST tests, a standard Toeplitz extraction transforms it into a final sequence that passes flawlessly. This outcome highlights a profound danger: since the validation process is insensitive to the quality of the raw data, it implies that even a fully predictable input could be processed to produce a certified, yet completely insecure, random sequence. Our theoretical analysis confirms this vulnerability extends to phase-noise-based QRNGs, suggesting a need for security validation to evolve beyond statistical analysis of the final output and consider the entire generation process.

quant-ph↗
arXiv2025

RL-Finetuned LLMs for Privacy-Preserving Synthetic Rewriting

The performance of modern machine learning systems depends on access to large, high-quality datasets, often sourced from user-generated content or proprietary, domain-specific corpora. However, these rich datasets inherently contain sensitive personal information, raising significant concerns about privacy, data security, and compliance with regulatory frameworks. While conventional anonymization techniques can remove explicit identifiers, such removal may result in performance drop in downstream machine learning tasks. More importantly, simple anonymization may not be effective against inference attacks that exploit implicit signals such as writing style, topical focus, or demographic cues, highlighting the need for more robust privacy safeguards during model training. To address the challenging issue of balancing user privacy and data utility, we propose a reinforcement learning framework that fine-tunes a large language model (LLM) using a composite reward function that jointly optimizes for explicit and implicit privacy, semantic fidelity, and output diversity. To effectively capture population level regularities, the privacy reward combines semantic cues with structural patterns derived from a minimum spanning tree (MST) over latent representations. By modeling these privacy-sensitive signals in their distributional context, the proposed approach guides the model to generate synthetic rewrites that preserve utility while mitigating privacy risks. Empirical results show that the proposed method significantly enhances author obfuscation and privacy metrics without degrading semantic quality, providing a scalable and model-agnostic solution for privacy preserving data generation in the era of large language models.

cs.CR↗
arXiv2025

Privacy-Preserving Synthetic Review Generation with Diverse Writing Styles Using LLMs

The increasing use of synthetic data generated by Large Language Models (LLMs) presents both opportunities and challenges in data-driven applications. While synthetic data provides a cost-effective, scalable alternative to real-world data to facilitate model training, its diversity and privacy risks remain underexplored. Focusing on text-based synthetic data, we propose a comprehensive set of metrics to quantitatively assess the diversity (i.e., linguistic expression, sentiment, and user perspective), and privacy (i.e., re-identification risk and stylistic outliers) of synthetic datasets generated by several state-of-the-art LLMs. Experiment results reveal significant limitations in LLMs' capabilities in generating diverse and privacy-preserving synthetic data. Guided by the evaluation results, a prompt-based approach is proposed to enhance the diversity of synthetic reviews while preserving reviewer privacy.

cs.CL↗
arXiv2025

On the rigidity of manifolds with respect to Gagliardo-Nirenberg inequalities

In this paper, we investigate local rigidity properties related to Gagliardo-Nirenberg constants and unweighted Yamabe-type constants. Let $V$ be an open bounded subset of an $n$-dimensional Riemannian manifold $(M,g)$ whose Gagliardo-Nirenberg constant satisfies \[ \mathbb{G}_{\alpha}^{\pm}(V,g) \geq \mathbb{G}_{\alpha}^{\pm}(\mathbb{R}^n,g_{\mathbb{R}^n}), \] where $(\mathbb{R}^n,g_{\mathbb{R}^n})$ denotes the $n$-dimensional Euclidean space with its standard metric. We show that for $\alpha \in (0,1) \cup \left(1,\frac{n+6}{n+2}\right)$ when $n \leq 6$ or $\alpha \in (0,1) \cup \left(1,\frac{n}{n-2}\right]$ when $n \geq 7$, if the first eigenvalue of the Ricci tensor satisfies \[ \int_V \lambda_1(\operatorname{Rc}) \, d\mu_g \geq 0, \] then $V$ must be flat. When $\alpha$ belongs to a specific subinterval around $1$ within the above range, $\mathbb{G}_{\alpha}^{\pm}(V,g) \geq \mathbb{G}_{\alpha}^{\pm}(\mathbb{R}^n,g_{\mathbb{R}^n})$ and the weaker curvature condition of the scalar curvature \[ \int_{V} \operatorname{Sc} \, d\mu_g \geq 0 \] already imply that $V$ is flat. Moreover, we prove that for $\alpha$ sufficiently close to 1, the condition \[ \mathbb{Y}_{\alpha}^{\pm}(V,g) \geq \mathbb{G}_{\alpha}^{\pm}(\mathbb{R}^n,g_{\mathbb{R}^n}) \] on the unweighted Yamabe-type constants guarantees the flatness of $V$.

math.DG↗
arXiv2025

LLMs are Frequency Pattern Learners in Natural Language Inference

While fine-tuning LLMs on NLI corpora improves their inferential performance, the underlying mechanisms driving this improvement remain largely opaque. In this work, we conduct a series of experiments to investigate what LLMs actually learn during fine-tuning. We begin by analyzing predicate frequencies in premises and hypotheses across NLI datasets and identify a consistent frequency bias, where predicates in hypotheses occur more frequently than those in premises for positive instances. To assess the impact of this bias, we evaluate both standard and NLI fine-tuned LLMs on bias-consistent and bias-adversarial cases. We find that LLMs exploit frequency bias for inference and perform poorly on adversarial instances. Furthermore, fine-tuned LLMs exhibit significantly increased reliance on this bias, suggesting that they are learning these frequency patterns from datasets. Finally, we compute the frequencies of hyponyms and their corresponding hypernyms from WordNet, revealing a correlation between frequency bias and textual entailment. These findings help explain why learning frequency patterns can enhance model performance on inference tasks.

cs.CL↗
arXiv2025

S2LPP: Small-to-Large Prompt Prediction across LLMs

The performance of pre-trained Large Language Models (LLMs) is often sensitive to nuances in prompt templates, requiring careful prompt engineering, adding costs in terms of computing and human effort. In this study, we present experiments encompassing multiple LLMs variants of varying sizes aimed at probing their preference with different prompts. Through experiments on Question Answering, we show prompt preference consistency across LLMs of different sizes. We also show that this consistency extends to other tasks, such as Natural Language Inference. Utilizing this consistency, we propose a method to use a smaller model to select effective prompt templates for a larger model. We show that our method substantially reduces the cost of prompt engineering while consistently matching performance with optimal prompts among candidates. More importantly, our experiment shows the efficacy of our strategy across fourteen LLMs and its applicability to a broad range of NLP tasks, highlighting its robustness

cs.CL↗
arXiv2025

MMLongBench: Benchmarking Long-Context Vision-Language Models Effectively and Thoroughly

The rapid extension of context windows in large vision-language models has given rise to long-context vision-language models (LCVLMs), which are capable of handling hundreds of images with interleaved text tokens in a single forward pass. In this work, we introduce MMLongBench, the first benchmark covering a diverse set of long-context vision-language tasks, to evaluate LCVLMs effectively and thoroughly. MMLongBench is composed of 13,331 examples spanning five different categories of downstream tasks, such as Visual RAG and Many-Shot ICL. It also provides broad coverage of image types, including various natural and synthetic images. To assess the robustness of the models to different input lengths, all examples are delivered at five standardized input lengths (8K-128K tokens) via a cross-modal tokenization scheme that combines vision patches and text tokens. Through a thorough benchmarking of 46 closed-source and open-source LCVLMs, we provide a comprehensive analysis of the current models' vision-language long-context ability. Our results show that: i) performance on a single task is a weak proxy for overall long-context capability; ii) both closed-source and open-source models face challenges in long-context vision-language tasks, indicating substantial room for future improvement; iii) models with stronger reasoning ability tend to exhibit better long-context performance. By offering wide task coverage, various image types, and rigorous length control, MMLongBench provides the missing foundation for diagnosing and advancing the next generation of LCVLMs.

cs.CV↗
arXiv2025

Advancing Antiferromagnetic Nitrides via Metal Alloy Nitridation

Nitride materials, valued for their structural stability and exceptional physical properties, have garnered significant interest in both fundamental research and technological applications. The fabrication of high-quality nitride thin films is essential for advancing their use in microelectronics and spintronics. Yet, achieving single-crystal nitride thin films with excellent structural integrity remains a challenge. Here, we introduce a straightforward yet innovative metallic alloy nitridation technique for the synthesis of stable single-crystal nitride thin films. By subjecting metal alloy thin films to a controlled nitridation process, nitrogen atoms integrate into the lattice, driving structural transformations while preserving high epitaxial quality. Combining nanoscale magnetic imaging with a diamond nitrogen-vacancy (NV) probe, X-ray magnetic linear dichroism, and comprehensive transport measurements, we confirm that the nitridated films exhibit a robust antiferromagnetic character with a zero net magnetic moment. This work not only provides a refined and reproducible strategy for the fabrication of nitride thin films but also lays a robust foundation for exploring their burgeoning device applications.

cond-mat.mtrl-sci↗
arXiv2025

Neutralizing Bias in LLM Reasoning using Entailment Graphs

LLMs are often claimed to be capable of Natural Language Inference (NLI), which is widely regarded as a cornerstone of more complex forms of reasoning. However, recent works show that LLMs still suffer from hallucinations in NLI due to attestation bias, where LLMs overly rely on propositional memory to build shortcuts. To solve the issue, we design an unsupervised framework to construct counterfactual reasoning data and fine-tune LLMs to reduce attestation bias. To measure bias reduction, we build bias-adversarial variants of NLI datasets with randomly replaced predicates in premises while keeping hypotheses unchanged. Extensive evaluations show that our framework can significantly reduce hallucinations from attestation bias. Then, we further evaluate LLMs fine-tuned with our framework on original NLI datasets and their bias-neutralized versions, where original entities are replaced with randomly sampled ones. Extensive results show that our framework consistently improves inferential performance on both original and bias-neutralized NLI datasets.

cs.CL↗