arXiv subjects

Shan Li

Publications and source records attributed to Shan Li.

At least 19 recordsLinked to original sources

arXiv2026

ContainmentBench: Trace-Based Evaluation of Post-Exposure Containment in Tool-Using LLM Agents

Tool-using large language model (LLM) agents read untrusted content, maintain memory, delegate tasks, and invoke tools with external side effects. Terminal attack-success or policy-violation rates do not show what happens between exposure and commit or whether a defense also suppresses authorized actions. We introduce ContainmentBench, a sandboxed benchmark comprising a 504-scenario specification dataset, a shared rollout-trace schema, and stage-scoped metrics for endpoint violations, logged propagation, and explicitly authorized taint-exposed proposals that commit. The main Qwen2.5-7B-Instruct study evaluates seven policy conditions and five seeds, yielding a 17,640-record trace corpus. Across 600 matched active-tainted rollout pairs, no committed policy violation was observed under either taint-only or intent-ledger enforcement. Their execution records nevertheless differed: 441 pairs (73.5%) had different values in a shared 12-field trace summary that includes commit-related diagnostics, and the mean authorized proposal-commit score was 0.164 under taint-only enforcement and 0.857 under intent-ledger enforcement, compared with 0.923 under tool-boundary enforcement. Logged-propagation rankings changed with stage selection and normalization. In a limited set of custom AgentDojo-native workflows, committed violations were observed without defense and were not observed under either evaluated defense. A separate 6,048-rollout Mistral/common-JSON model-interface configuration retained the v1-to-v2 proposal-commit improvement, but committed violations were observed under intent-ledger v2. Equal terminal outcomes do not imply equal containment. The evaluation uses synthetic workflows. The intended intent-ledger mechanism assumes schema-aligned authorization metadata; one public-status task family violates this assumption and is analyzed separately.

cs.CR↗

arXiv2026

From Refusal Geometry to Safety Geometry: Harmfulness--Refusal Coupling under Dynamic Adversarial Fine-Tuning

Safety alignment requires language models to refuse harmful requests without losing the ability to answer benign ones. Existing robustness evaluations, however, do not reveal whether a model has learned to recognize harmfulness, to activate a refusal policy, or to couple these two processes. We study this question with a dual safety-geometry protocol that measures harmfulness carriers, refusal carriers, and their coupling across aligned instruction-tuned anchors and matched Mistral-7B-v0.1 SFT/R2D2 training trajectories. The aligned anchors validate the protocol: refusal-side interventions reopen attack success more strongly than harmfulness-only interventions, while harmfulness and refusal carriers remain nearly orthogonal. Along the Mistral trajectory, R2D2 exhibits a high-coupling early phase with strong fixed-source robustness, saturated safe-prompt refusal, and collapsed benign utility. Later checkpoints move to a lower-coupling regime with partial utility recovery and reopened attack success. SFT provides an important contrast: it also reaches low coupling, but remains substantially less robust, showing that low coupling alone is not a safety guarantee. All-anchor diagnostics and sparse GCG/AutoDAN transfer experiments further show that H/R coupling is informative in the R2D2 regime, whereas SFT transfer is better summarized by drift or behavior-state measures. Causal sweeps support fixed-protocol sensitivity relative to matched unit-direction controls, but do not establish independent harmfulness and refusal pathways. These results frame harmfulness--refusal coupling as an operational diagnostic for safety-geometry dynamics under adversarial fine-tuning.

cs.CR↗

arXiv2026

An AI-driven robotic system for two-dimensional hetero-assemblies

Nanomaterials stacked on-demand, such as rotationally assembled two-dimensional (2D) van der Waals (vdW) layered compounds, provides a versatile platform for quantum simulation and the exploration of exotic electronic phases. Currently, however, such nanoassemblies remain largely confined to inefficiency, manually operated process, limiting their potential for probing emergent physical phenomena. There is a pressing need in the field for high-precision, automated assembling techniques, especially for the scalable fabrication of 2D twistronic heterostructures. Here, we present an intelligent automation system dedicated to the fabrication of van der Waals stacks, following the state-of-the-art protocol for dry transfer of exfoliated 2D materials. The system further employs metadata generated from each automated stacking procedure to perform reinforcement learning, thereby continuously bettering its performances. As a concrete demonstration, we fabricate twisted bilayer graphene (TBLG) -- known for its challenging preparation -- and exhibit its unconventional superconductivity near the magic angle. Our work may pave the way for high-throughput fabrication of low-dimensional nanomaterials including twistronic heterostructures, where integrating data mining and artificial intelligence can accelerate the discovery of novel physical phenomena.

cond-mat.mes-hall↗

arXiv2026

Constraining the magnetic field strength of a flaring radio core in the compact steep spectrum source 3C 138

Compact steep spectrum (CSS) sources generally show weak Doppler boosting, yet some exceptions show multi-year-scale radio flux variability and high-energy activity. Since 2022, the CSS quasar 3C 138 has been in a radio high state accompanied by multiple gamma-ray outbursts, offering unique opportunities to study changes in jet physical conditions. We estimated the synchrotron self-absorption (SSA) magnetic field ($B_{\rm SSA}$) in the SSA core of 3C 138 during its high state and compared it with the equipartition magnetic field ($B_{\rm eq}$) to assess the core field environment. Using extended Korean Very long-baseline interferometry Network (KVN) data at 22, 43, 86, and 129 GHz (2024-2025), we calibrated the visibilities and modeled resolved components with circular Gaussians. A single-zone SSA model fitted to the core spectrum provided the turnover frequency and peak flux density, from which we estimated the $B_{\rm SSA}$ and $B_{\rm eq}$. We used Very Large Array and Atacama Large Millimeter/submillimeter Array data to constrain the broadband spectra with the same model. The KVN SSA core shows a turnover at about 33 GHz and a peak flux of about 1.45 Jy. The inferred $B_{\rm SSA}$ is far below equipartition, with $B_{\rm SSA}/B_{\rm eq}\approx0.05$. The flux variability of 3C 138 is driven by a compact, particle-dominated core. Shock-driven particle injection in the inner jet could account for the core brightening and the production of X-ray/gamma-ray emissions through an inverse-Compton process without requiring extreme relativistic beaming effects.

astro-ph.GA↗

arXiv2026

Generativism: Toward a Learning Theory for the Age of Generative Artificial Intelligence

The four dominant learning theories of behaviorism, cognitivism, constructivism, and connectivism show significant conceptual limitations as generative artificial intelligence (AI) proliferates in educational settings. These frameworks were formulated before the emergence of AI systems capable of generating, synthesizing, and reasoning about knowledge. This article critically examines each learning theory and identifies assumptions challenged by the affordances of generative AI. Drawing on research in distributed cognition, extended mind, human-AI collaboration, AI literacy, cognitive offloading, and metacognition, the article proposes Generativism as a learning theory for the generative AI age. Generativism posits that learning increasingly occurs through the iterative co-construction of knowledge between human learners and AI systems. The proposed framework is organized around four constructs (epistemic partnership, distributed agency, generative literacy, and adaptive metacognition) and four principles specifying the relations among them. The framework offers a foundation for rethinking instructional design, learning, assessment, and expertise development in contexts where generative AI plays an integral role in cognition.

cs.CY↗

arXiv2026

Dynamic Adversarial Fine-Tuning Reorganizes Refusal Geometry

Safety-aligned language models must refuse harmful requests without broad over-refusal, but it remains unclear how dynamic adversarial fine-tuning changes refusal-control carriers: Kullback--Leibler (KL)-constrained directions or small subspaces that causally modulate refusal without large safe-prompt distribution shifts. We study a 7B backbone under supervised fine-tuning (SFT) and Robust Refusal Dynamic Defense (R2D2), aligning HarmBench, StrongREJECT, and XSTest evaluations with five-anchor geometry measurements, causal interventions, and sparse adaptive stress tests. R2D2 drives fixed-source HarmBench attack success to zero at early checkpoints; however, these checkpoints also exhibit maximal XSTest refusal and fail a benign-utility audit. Later checkpoints partially recover utility-facing behavior while reopening attack success, with adaptive GCG attack success rate rising to 0.415 at step 250 and 0.613 at step 500. Internally, R2D2 preserves a late-layer admissible refusal-control carrier through step 100 and then relocates the best admissible carrier to an early layer; SFT relocates earlier yet remains less robust. Effective rank stays near 1.24, and SFT shows larger principal-angle drift, arguing against both dimensional expansion and drift magnitude as sufficient explanations. Causal interventions support a low-dimensional but utility-coupled carrier. These results support a geometry-reorganization account of R2D2 along a robustness--utility frontier, without establishing adaptive robustness.

cs.LG↗

arXiv2026

A Scoping Review of Large Language Model-Based Pedagogical Agents

This scoping review examines the emerging field of Large Language Model (LLM)-based pedagogical agents in educational settings. While traditional pedagogical agents have been extensively studied, the integration of LLMs represents a transformative advancement with unprecedented capabilities in natural language understanding, reasoning, and adaptation. Following PRISMA-ScR guidelines, we analyzed 52 studies across five major databases from November 2022 to January 2025. Our findings reveal diverse LLM-based agents spanning K-12, higher education, and informal learning contexts across multiple subject domains. We identified four key design dimensions characterizing these agents: interaction approach (reactive vs. proactive), domain scope (domain-specific vs. general-purpose), role complexity (single-role vs. multi-role), and system integration (standalone vs. integrated). Emerging trends include multi-agent systems that simulate naturalistic learning environments, virtual student simulation for agent evaluation, integration with immersive technologies, and combinations with learning analytics. We also discuss significant research gaps and ethical considerations regarding privacy, accuracy, and student autonomy. This review provides researchers and practitioners with a comprehensive understanding of LLM-based pedagogical agents while identifying crucial areas for future development in this rapidly evolving field.

cs.AI↗

arXiv2026

Where within the 3C 84 jet are $\gamma$-rays produced?

The location of $\gamma$-ray creation and emission within extra-galactic jets is a matter of active debate. One particularly well-suited source to pinpoint the location is the nearby, bright radio galaxy 3C 84, harbouring a powerful jet. Here we investigate the origin of $\gamma$-rays measured during a recent $\gamma$-ray flare, by analysing the linear polarisation signal of close-in-time very long baseline interferometry (VLBI) observations at centimetre and millimetre wavelengths. While 3C 84 is overall almost unpolarised, we find that close-in-time to the $\gamma$-ray flare peak regions at parsec-scale distances from the central engine shows a fractional linear polarisation increase. Under the physically well-motivated assumption of a causal relation between this polarisation enhancement and the $\gamma$-ray flare, and combined with insights from concurrent X-ray polarisation measurements, the $\gamma$-rays being created in this region is a physically motivated scenario, in a process consistent with synchrotron self-Compton.

astro-ph.HE↗

arXiv2026

Sharp estimates for the Robin Laplacian under a perimeter constraint in hyperbolic space

In this paper, we establish a lower bound, in terms of the isoperimetric deficit, for the first eigenvalue of the Robin Laplacian with negative boundary parameter on horospherically convex bounded domains in the hyperbolic space. This implies that the geodesic ball maximizes this eigenvalue among all such domains, thereby providing a partial resolution to an open problem posed by Celentano, Krej\v{c}i\v{r}\'{i}k and Lotoreichik in \cite{CKL26}. Furthermore, we derive upper bounds for the first eigenvalue of the Robin Laplacian with positive boundary parameter on horospherically convex bounded domains in the hyperbolic space.

math.DG↗

arXiv2026

Investigating Self-regulated Learning Sequences within a Generative AI-based Intelligent Tutoring System

There has been a growing trend in employing generative artificial intelligence (GenAI) techniques to support learning. Moreover, scholars have reached a consensus on the critical role of self-regulated learning (SRL) in ensuring learning effectiveness within GenAI-assisted learning environments, making it essential to capture students' dynamic SRL patterns. In this study, we extracted students' interaction patterns with GenAI from trace data as they completed a problem-solving task within a GenAI-assisted intelligent tutoring system. Students' purpose of using GenAI was also analyzed from the perspective of information processing, i.e., information acquisition and information transformation. Using sequential and clustering analysis, this study classified participants into two groups based on their SRL sequences. These two groups differed in the frequency and temporal characteristics of GenAI use. In addition, most students used GenAI for information acquisition rather than information transformation, while the correlation between the purpose of using GenAI and learning performance was not statistically significant. Our findings inform both pedagogical design and the development of GenAI-assisted learning environments.

cs.CY↗

arXiv2025

Detection of Compton scattering in the jet of 3C 84

3C 84 is the brightest cluster galaxy in the Perseus Cluster. It is among the closest radio-loud active galaxies and among the very few that can be detected from low frequency radio up to TeV $\gamma$-rays. Here we report on the first X-ray polarization observation of 3C~84 with the Imaging X-ray Polarimetry Explorer, for a total of 2.2 Msec that coincides with a flare in $\gamma$-rays. This is the longest observation for a radio-loud active galaxy that allowed us to reach unprecedented sensitivity, leading to the detection of an X-ray polarization degree of $\rm\Pi_X=4.2\pm1.3\%$ ($\sim3.2\sigma$ confidence) at an X-ray electric vector polarization angle of $\rm \psi_X=163^{\circ}\pm9^{\circ}$, that is aligned with the radio jet direction on the sky. Optical polarization observations show fast variability about the jet axis as well. Our results strongly favor models in which X-rays are produced by Compton scattering from relativistic electrons -- specifically Synchrotron Self-Compton -- that takes places downstream, away from the supermassive black hole.

astro-ph.HE↗

arXiv2025

Unveiling blazar synchrotron emission: a multiwavelength polarimetric study of HSP and LSP populations

Polarimetric properties of blazars allow us to put constraints on the acceleration mechanisms that fuel their powerful jets. By studying the multiwavelength polarimetric behaviour of high synchrotron peaked (HSP) and low synchrotron peaked (LSP) blazars, we aim to explore differences in their emission mechanisms and magnetic field structure in the acceleration region. In this study, we take advantage of several X-ray polarisation observations of HSP by the IXPE, including four new observations of Mrk 501, and optical polarisation observations of LSP from RoboPol and many others. We find that the polarisation degree (PD) distribution of HSP in X-rays is systematically higher than in optical and mm-radio wavelengths, as reported in previous IXPE publications. The distribution of the X-ray electric vector position angles (PA) is centered around the jet axis with most of the observations consistent with zero difference within uncertainties. In fact, the distribution of the offset of the PA from the jet axis is consistent between the LSP and HSP populations (with PA measured in optical for the first, X-ray for the latter), suggesting a common magnetic field structure close to the acceleration region, in strong support of the emerging energy stratified picture of particle acceleration followed by energy loss in blazar jets.

astro-ph.HE↗

arXiv2025

A Theory-driven and AI-enhanced Simulation Platform for Cultivating Nutrition Literacy

This study introduces and evaluates Healthy Choice, an innovative theory-driven and AI-enhanced simulation platform designed to cultivate nutrition literacy through interactive scenario-based learning experiences. We collected feedback from 114 university students with diverse backgrounds who completed simulated product selection scenarios. Quantitative ratings of usefulness and ease of use demonstrated high user satisfaction.

cs.HC↗

arXiv2025

On the Bossel-Daners inequality for the p-Laplacian on complete Riemannian manifolds

In this paper, we obtain the Bossel-Daners inequality for the first eigenvalue of the p-Laplacian with Robin boundary conditions on complete Riemannian manifolds with lower Ricci curvature bounds. Furthermore, we demonstrate that the Bossel-Daners inequality extends to compact submanifolds within complete Riemannian manifolds characterized by positive asymptotic volume ratio and non-negative intermediate Ricci curvature.

math.DG↗

arXiv2024

Frequency-based Matcher for Long-tailed Semantic Segmentation

The successful application of semantic segmentation technology in the real world has been among the most exciting achievements in the computer vision community over the past decade. Although the long-tailed phenomenon has been investigated in many fields, e.g., classification and object detection, it has not received enough attention in semantic segmentation and has become a non-negligible obstacle to applying semantic segmentation technology in autonomous driving and virtual reality. Therefore, in this work, we focus on a relatively under-explored task setting, long-tailed semantic segmentation (LTSS). We first establish three representative datasets from different aspects, i.e., scene, object, and human. We further propose a dual-metric evaluation system and construct the LTSS benchmark to demonstrate the performance of semantic segmentation methods and long-tailed solutions. We also propose a transformer-based algorithm to improve LTSS, frequency-based matcher, which solves the oversuppression problem by one-to-many matching and automatically determines the number of matching queries for each class. Given the comprehensiveness of this work and the importance of the issues revealed, this work aims to promote the empirical study of semantic segmentation tasks. Our datasets, codes, and models will be publicly available.

cs.CV↗

arXiv2024

CloudFort: Enhancing Robustness of 3D Point Cloud Classification Against Backdoor Attacks via Spatial Partitioning and Ensemble Prediction

The increasing adoption of 3D point cloud data in various applications, such as autonomous vehicles, robotics, and virtual reality, has brought about significant advancements in object recognition and scene understanding. However, this progress is accompanied by new security challenges, particularly in the form of backdoor attacks. These attacks involve inserting malicious information into the training data of machine learning models, potentially compromising the model's behavior. In this paper, we propose CloudFort, a novel defense mechanism designed to enhance the robustness of 3D point cloud classifiers against backdoor attacks. CloudFort leverages spatial partitioning and ensemble prediction techniques to effectively mitigate the impact of backdoor triggers while preserving the model's performance on clean data. We evaluate the effectiveness of CloudFort through extensive experiments, demonstrating its strong resilience against the Point Cloud Backdoor Attack (PCBA). Our results show that CloudFort significantly enhances the security of 3D point cloud classification models without compromising their accuracy on benign samples. Furthermore, we explore the limitations of CloudFort and discuss potential avenues for future research in the field of 3D point cloud security. The proposed defense mechanism represents a significant step towards ensuring the trustworthiness and reliability of point-cloud-based systems in real-world applications.

cs.CV↗

arXiv2024

A Multi-wavelength Study on A Gamma-ray Bright AGN 1308+326 Using KVN at 22 and 43 GHz

In this paper, we conduct a multi-frequency analysis of the gamma-ray bright blazar 1308+326 from February 2013 to March 2020, using the Korean VLBI Network at 22 and 43 GHz and gamma-ray data from the Fermi Large Area Telescope (LAT). Our findings reveal spectral variations around the 2014 gamma-ray flare, aligning with the shock-in-jet model. A strong correlation is observed between gamma-ray and 43 GHz emissions, with a 27-day lag in the VLBI core light curve, indicating a 50-day delay from the beginning of a specific radio flare to the gamma-ray peak. This radio flare correlates with a new jet component, suggesting the 2014 gamma-ray flare resulted from its interaction with a stationary component. Our analysis indicates the 2014 gamma-ray flare originated 40-63 parsecs from the central engine, with seed photons for the gamma-ray emission unlikely from the broad-line region.

astro-ph.HE↗

arXiv2024

PikeLPN: Mitigating Overlooked Inefficiencies of Low-Precision Neural Networks

Low-precision quantization is recognized for its efficacy in neural network optimization. Our analysis reveals that non-quantized elementwise operations which are prevalent in layers such as parameterized activation functions, batch normalization, and quantization scaling dominate the inference cost of low-precision models. These non-quantized elementwise operations are commonly overlooked in SOTA efficiency metrics such as Arithmetic Computation Effort (ACE). In this paper, we propose ACEv2 - an extended version of ACE which offers a better alignment with the inference cost of quantized models and their energy consumption on ML hardware. Moreover, we introduce PikeLPN, a model that addresses these efficiency issues by applying quantization to both elementwise operations and multiply-accumulate operations. In particular, we present a novel quantization technique for batch normalization layers named QuantNorm which allows for quantizing the batch normalization parameters without compromising the model performance. Additionally, we propose applying Double Quantization where the quantization scaling parameters are quantized. Furthermore, we recognize and resolve the issue of distribution mismatch in Separable Convolution layers by introducing Distribution-Heterogeneous Quantization which enables quantizing them to low-precision. PikeLPN achieves Pareto-optimality in efficiency-accuracy trade-off with up to 3X efficiency improvement compared to SOTA low-precision models.

cs.LG↗