arXiv subjects

Dan Li

Publications and source records attributed to Dan Li.

At least 19 recordsLinked to original sources

arXiv2026

The Surprising Effectiveness of LLMs in BGP Security: Mining An Unprecedented Amount of Incidents and Boosting Anomaly Detection

Border Gateway Protocol (BGP) security is critical to Internet infrastructure, yet progress in routing anomaly detection has been limited by the scarcity of publicly available incident datasets, which contain only 18 recorded cases. We observe that public operator mailing lists, e.g., NANOG and AusNOG, contain abundant yet largely untapped reports of real-world routing anomalies. To leverage this source, we develop an LLM-assisted extraction pipeline that identifies 244 candidate incidents from historical discussion threads. After expert validation, we curate a verified benchmark containing 232 confirmed routing anomaly events, making it 11.89X larger than existing dataset. Using this benchmark, we show that existing routing anomaly detection systems generalize poorly to diverse real-world incidents. At the same time, we find that some general-purpose LLMs without routing-specific adaptation can identify a subset of routing anomalies, but their performance varies across models and remains insufficient for reliable routing anomaly detection. Motivated by this observation, we design ROUTELLM, an LLM-based routing anomaly detector that integrates BGP-semantic-aware tokenization, routing-domain adaptation, and time-aware routing evidence retrieval. Experimental results show that ROUTELLM achieves 87.13% event-level accuracy and 94.65% message-level accuracy, outperforming the strongest baselines by 55.30% and 68.50%, respectively. We open-source the verified routing anomaly benchmark, fine-tuned model, and implementation code to support future research on BGP security.

cs.NI↗

arXiv2026

Anthropogenic Heat in Urban Climate Systems: Forcing, Sensitivity, and Feedback

Anthropogenic heat flux, the heat released to the environment from human activities such as building energy use, transportation, industrial processes, and human metabolism, is a defining feature of the urban climate system. It is an important contributor to the urban heat island (UHI) effect and influences a wide range of urban meteorological processes. Its significance extends beyond urban climatology because of its close connections to energy consumption, greenhouse gas emissions, and climate adaptation. Despite decades of research, anthropogenic heat flux remains one of the least well-constrained components of the urban energy balance. Moreover, its climatic significance has often been assessed from an applied perspective, with less emphasis on developing transferable physical understanding of how anthropogenic heat flux acts as a forcing, how urban temperatures respond, and how feedbacks modify that response. This review develops a forcing-response-feedback framework for synthesizing current understanding of the role of anthropogenic heat flux in the urban climate system and identifies priorities for future research.

physics.ao-ph↗

arXiv2026

The electrical transport of intrinsic two-dimensional ferroelectric metal PtBi2

Breaking the conventional stereotype that ferroelectrics are necessarily insulating, two-dimensional (2D) ferroelectric metals combine seemingly incompatible switchable electric polarization and metallic conductivity, providing a fertile ground for the discovery of novel electrical transport phenomena and the development of innovative electronic devices. Using the semiclassical Boltzmann equation and first-principles calculations, we systematically investigate the linear and nonlinear transport responses of the intrinsic 2D ferroelectric metal \ch{PtBi2} to an applied electric field. Our \textit{ab initio} molecular dynamics simulations reveal that it possesses a high Curie temperature reaching $800~\text{K}$. We propose that the crystal structure of its high-temperature paraelectric phase can be explicitly distinguished through simple measurements of the in-plane electrical conductivity. Quantitative calculations of the Edelstein effect and the intrinsic spin Hall effect demonstrate a sizable charge-to-spin conversion efficiency, highlighting its potential in spintronics. We also find that a Berry curvature dipole-induced nonlinear Hall effect emerges in uniaxially strained \ch{PtBi2}. Furthermore, we highlight the unique advantages of 2D ferroelectric metals in gate-controlled transport applications. Based on the domain wall scattering mechanism, we conceptually design a novel ferroelectric metal field-effect transistor (FEM-FET) capable of nonvolatile switching between high-resistance and low-resistance states under a gate voltage. Our work not only unveils the rich transport physics in 2D ferroelectric metals but also provides valuable insights into the design of next-generation nonvolatile memory and spintronic devices.

cond-mat.mtrl-sci↗

arXiv2026

TideRL: Boosting Agentic RL Goodput with Readiness-Aware Scheduling

Reinforcement learning (RL) for large language models is moving toward multi-turn agentic workloads, where rollout tasks repeatedly pause for external environments, resume with growing contexts, and finish at highly variable times. In this setting, RL training goodput, measured by training throughput, matters more than raw GPU occupancy: GPU waiting and repeated prefill recomputation are pure overhead. We present TideRL, a readiness-aware elastic RL system with Continuous Task Batching, Resource-Aware Ref-Actor Pipelining, and Elastic Resource Scaling. CTB preserves useful rollout state, $\textrm{RA}^2\textrm{P}$ selects between decoupled streaming and colocated aggregation from the ready backlog and arrival interval, and ERS moves ranks between rollout and training using the same readiness signals. Across text-only and multi-modal agentic workloads, TideRL improves RL training goodput by up to 5.6$\times$ over synchronous baselines and over 33% over asynchronous baselines, while reaching similar task performance. It also improves KV cache hit rate by 1.58$\times$, reduces per-step training time by up to 44.3%, and cuts total waiting time by up to 77.6%.

cs.LG↗

arXiv2026

Construction of an analytic multi-component accretion environment and its application to Kerr black hole imaging

The construction of accretion environments is fundamental to black hole imaging. From a purely geometric perspective, we construct a novel analytic accretion environment comprising a geometrically thick disk, ring-like bumps with a Gaussian profile, and localized compact emission regions modeled by Gaussian distributions. This environment offers high flexibility, enabling independent adjustments of disk thickness, vertical structure, and the positions and morphologies of localized spots, thereby allowing it to qualitatively mimic high-energy astrophysical phenomena. Applying this model to the Kerr spacetime, we investigate the resulting images via radiative transfer and ray-tracing simulations. The results validate the effectiveness of our accretion model and reveal novel observational signatures of Kerr black holes under multi-component illumination, including multiple bright spots and arc-like structures. This work provides a convenient and fully analytic framework for modeling accretion in curved spacetimes, and offers a new perspective on inferring accretion mechanisms and transient high-energy processes from image features.

gr-qc↗

arXiv2026

An Empirical Study of Coordination Mode as the First-Class Citizen in From-Scratch Multi-Agent Coding

Multi-agent vibe coding promises to accelerate software development, yet existing benchmarks rely on synthetic environments that ignore practical time and monetary costs, conflate reasoning with communication, and reward only superficial completion. We introduce multi-agent from-scratch evaluation benchmark, MSEval, evaluating multi-agent coding on real-world tasks. Grounded in 10 authentic, full-stack projects across 10 domains, MSEval scores performance using hierarchical requirements and deterministic rubrics. Its execution engine, LegoGent, tests 10 collaboration topologies where agents coordinate via periodic sync intervals and deploy through native CI/CD pipelines. Concurrently, the automated grader TAgent dynamically probes implementations to jointly measure functional success, latency, and prefix-cached token cost. Across 100 runs, MSEval reveals that organizational topology rivals model capability in shaping the speed--cost--quality trade-off. For identical tasks and models, varying the topology shifts scores by over 30 points and doubles wall-clock time. Structured pipelines converge fastest with the highest quality, whereas heavy managerial oversight degrades performance. Ultimately, MSEval establishes a rigorous, reproducible standard for measuring how multi-agent teams actually build software. The benchmark is released at https://github.com/robinren03/MSEval.

cs.AI↗

arXiv2026

Factorized Spectral Representations for Reinforcement Learning

Learning a compact model of the world from interaction data is central to sample-efficient deep reinforcement learning. Spectral representation methods have become the leading paradigm for representation learning in continuous control by taking a matrix view of the transition kernel, with state-action pairs on one side and next states on the other, and learning a low-rank factorization through self-supervised contrastive objectives. We take this view one step further. The transition kernel is naturally a three-mode tensor over states, actions, and next states, and a CP decomposition gives one feature map per mode. We propose FaStR, which fits this decomposition with a noise contrastive objective, producing separate state, action, and next-state encoders that together form a single spectral representation. The factored form yields a smaller hypothesis class, and the sample size needed for representation learning shrinks by a factor that scales with the smaller of the state and action dimensions. Empirically, FaStR delivers its largest gains on high-dimensional locomotion tasks whose dynamics align with the factored structure, and the learned state encoder transfers intact across actuator shift while only the action encoder is retrained.

cs.LG↗

arXiv2026

Spectral extremal problems on planar and outerplanar graphs without $C_{k,l}

Let $\emph{spex}_{\mathcal{P}}(n,F)$ and $\emph{spex}_{\mathcal{OP}}(n,F)$ be the maximum spectral radius among all $n$-vertex $F$-free planar graphs and outerplanar graphs, respectively. Define $C_{k,l}$ as a graph obtained from $C_k \cup C_l$ such that the two cycles share a common vertex, where $l \ge k \ge 3$. In the 1990s, Cvetkovi\'c and Rowlinson conjectured $K_1 + P_{n-1}$ maximizes spectral radius in outerplanar graphs on $n$ vertices, while Boots and Royle (independently, Cao and Vince) conjectured $K_2 + P_{n-2} $ does so in planar graphs. Tait and Tobin [J. Combin. Theory Ser. B, 2017] determined the fundamental structure as the key to confirming these two conjectures for sufficiently large $n$. Recently, Yin and Li [Discrete Mathematics, 2026] characterized the extremal graphs for $\emph{spex}_{\mathcal{P}}(n,B_{t,l})$ and $\emph{spex}_{\mathcal{OP}}(n,B_{t,l})$ in planar and outerplanar graphs on the basis of this key idea, where $B_{t,l}$ denotes the graph obtained by $t$ edge-disjoint $l$-cycles sharing a common vertex. In this paper, we focus on planar and outerplanar graphs without $C_{k,l}$, and determine $\emph{spex}_{\mathcal{P}}(n,C_{k,l})$ and $\emph{spex}_{\mathcal{OP}}(n,C_{k,l})$ along with their unique extremal graphs for all $l \geq k \geq 3$ and large $n$.

math.CO↗

arXiv2026

AdaPCLA: Adaptive Prior-Calibrated Logit Adjustment for Long-Tailed Longitudinal EHR Generation

Generative modeling of longitudinal Electronic Health Records is increasingly important for privacy-preserving research, yet standard autoregressive models tend to underrepresent the co-occurrence structure of tail events (i.e., diseases, symptoms), reducing the fidelity and faithfulness of generated data for rare subpopulations. To this end, we propose AdaPCLA framework, which enables generative models to adaptively fit and generate EHR data through a data distribution-aware training strategy; this is achieved by internalizing data knowledge parameters by simulated annealing training. It also supports training-free adaptation to a diverse clinical population for generation through zero-shot distribution control. Moreover, our theoretical analysis characterizes rare-code logit updates through the label-wise empirical NTK and derives a prior-internalization bound for how annealing speed and NTK conditioning affect retained prior signals. Experiments on real-world data show that AdaPCLA achieves consistent gains in tail plausibility, downstream utility, and zero-shot control; in particular, it improves TailPairSeen over HALO by 114.2% on MIMIC-III and 65.1% on MIMIC-IV, outperforms GPT-style generation by 3.5% F1 for zero-shot cross-population adaptation.

cs.LG↗

arXiv2026

Signal-Guided Optimization for Machine Unlearning

Current machine unlearning methods predominantly rely on global, coarse-grained intervention strategies. They lack precise pilot signals to guide the unlearning process and fail to provide differentiable guidance across different unlearning tasks. Due to the varying memorization strengths of samples during original training, such a uniform strategy leads to two problems: some samples are over-unlearned, which harms model utility; while others are under-unlearned, leaving residual information that can be exploited by privacy attacks. In this paper, we propose GSUO, a guidance-signal-aware unlearning optimization framework that designs task-specific fine-grained guidance signals to steer the unlearning process and is applicable to both random-subset and class-wise forgetting tasks. Extensive experiments demonstrate that GSUO outperforms 14 baselines in terms of both unlearning effectiveness and generalization, while achieving high efficiency and significant speedups, validating its effectiveness for reliable machine unlearning.

cs.LG↗

arXiv2026

Rational Orbits and Gravitational Waves in Static Spherical Spacetimes: An Open-Source Numerical Framework

Timelike orbits constitute a crucial probe for exploring the intrinsic properties of curved spacetimes, and the carried gravitational radiation signals provide a direct window into strong field gravity. In this paper, we develop a versatile computational framework based on Mathematica and the OpenMP parallel architecture to simulate the rational orbits of timelike particles and their gravitational radiation in static spherically symmetric spacetimes. Specifically, requiring only the user defined covariant metric, this numerical tool can efficiently calculate rational orbits across various configurations, as well as the corresponding gravitational wave polarization states and characteristic strains. The package presented here offers a highly efficient and comprehensive one-stop solution for investigating the properties of curved spacetimes and their potential observational signatures. To demonstrate the reliability and capability of our code, we apply it to the Schwarzschild spacetime as a test case, illustrating the functionality of the code across several key aspects, including the effective potential, stable orbital regions, rational and irrational orbits, and gravitational wave signals. Furthermore, we show that the gravitational waves emitted by an extreme-mass-ratio inspiral system composed of an intermediate mass black hole and the Galactic Center supermassive black hole have the potential to be identified by future space detectors.

gr-qc↗

arXiv2026

Binary Decompilation LLM with Feedback-Driven Multi-Turn Refinement

Binary decompilation is fundamental to security tasks such as vulnerability discovery, malware inspection, and executable-only program understanding. Recent LLM-based decompilation methods have shown promising results, but most still follow a single-turn generation paradigm: given assembly code or decompiler-produced pseudo-code, the model generates one output and stops. Consequently, the generated code may appear readable or even compile successfully, yet still deviate from the behavior of the original binary and mislead downstream analysis. This paper presents AutoDecompiler, a decompilation-specialized LLM trained with reinforcement learning for feedback-driven multi-turn binary decompilation. Instead of treating decompilation as one-shot code generation, AutoDecompiler formulates it as an iterative refinement process, where the model revises generated code based on compilation, execution, and input/output testing feedback. To enable this process, we design decompilation-specific rewards that capture code validity, recompilability, execution consistency, and semantic fidelity. We further construct stage-aware diagnostic feedback from compiler errors, execution failures, and failed test cases, and introduce progress-aware trajectory rewarding and turn-aware advantage reweighting to encourage beneficial revisions while suppressing regressions. We train the AutoDecompiler family and evaluate it across different input settings, model scales, and benchmarks. Experimental results show that AutoDecompiler consistently outperforms its single-turn counterparts under the same model size and input setting, achieving clear improvements in behavioral re-executability. These results demonstrate that learning to exploit program feedback with reinforcement learning is an effective direction for improving the functional correctness of LLM-based binary decompilation.

cs.SE↗

arXiv2026

Autonomous End-to-End SOH Prediction Services for Battery Systems via Temporal-Contrastive Representation Learning

Accurate state of health (SOH) estimation is a critical diagnostic service for lithium-ion battery management. However, reliance on labor-intensive manual feature engineering and opaque black-box models hinders scalable industrial deployment. To address this, we introduce TC-SOH: a modular, plug-and-play service architecture for autonomous, end-to-end SOH prediction. TC-SOH employs a temporal-contrastive mechanism and a cross-window prediction pretext task to extract degradation-relevant representations directly from raw operational data. To improve transparency, we connect model efficacy with representation diagnostics: visualization, sensitivity analysis, redundancy analysis, bidirectional probing, future-SOH probing, and temporal shuffling show that learned features overlap with selected expert descriptors while retaining additional SOH-relevant variation, and that ordered temporal context improves subsequent-SOH prediction. Across four public datasets, TC-SOH outperforms the considered physics-informed and data-driven baselines, reducing MAPE by 1.91 times and RMSE by 2.13 times.

cs.LG↗

arXiv2026

Activation Steering Induces Emergent Misalignment: A More Comprehensive Evaluation

Activation steering has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs). By constructing a steering vector from examples of a target behavior and injecting it into intermediate activations during inference, activation steering enables flexible behavioral control while avoiding the permanent parameter updates required by finetuning. Meanwhile, recent work has identified emergent misalignment (EM) as a significant safety concern, wherein models finetuned on unsafe examples from a narrow task may unexpectedly generalize to broadly unsafe behavior on unrelated tasks. Although finetuning-induced EM has been extensively studied, whether activation steering can induce EM remains comparatively under-explored, despite its increasing use as a model-control technique. In this paper, we present a comprehensive study of activation-steering-induced emergent misalignment, substantially expanding the evaluation scope beyond existing pioneering work. First, we show that activation steering can induce broad misalignment, even in the recent Qwen-3.5 series. Moreover, activation-steered models produce harmful responses with stronger semantic relevance and higher coherence than their finetuned counterparts, making the resulting misalignment potentially more harmful. Second, we characterize properties of AS-induced EM by analyzing key steering-specific factors, including steering magnitude, the low-rank structure of the steering subspace, and the number of epochs during steering-vector construction. Third, we evaluate the robustness and sensitivity of AS-induced EM across diverse model families, model scales, target tasks, and intervention layers. Our findings reveal activation steering as a significant yet under-examined source of emergent misalignment and provide an activation-space perspective for understanding the mechanisms and safety risks of EM.

cs.LG↗

arXiv2026

Exploiting Verification-Generation Gap: Test-Time Reinforcement Learning with Confidence-Conditioned Verification

Test-time reinforcement learning has emerged as a promising paradigm for enhancing the complex reasoning abilities of large language models in a completely label-free manner. Despite existing studies focusing on Pass@1 performance, optimizing Pass@k remains under-explored yet critical in label-free settings, which measures generation coverage for sustained exploration. Optimizing Pass@k in label-free setting is highly non-trivial, as directly applying the Pass@k advantage designs effective for RLVR yields unsatisfactory performance. Through in-depth empirical analysis, we discover the root causes hindering performance: pseudo-label estimations for low-confidence samples have a high probability of being incorrect, while candidate answers for high-confidence samples suffer from severe diversity collapse. To overcome these hurdles, we propose TTRL-CoCoV (Test-Time Reinforcement Learning with Confidence-Conditioned Verification), a novel confidence-adaptive framework that expands Pass@k coverage and improves Pass@1 performance. Based on our key insight that verification capability generally leads generation capability, TTRL-CoCoV employs a confidence-conditioned mechanism: for high-confidence samples, it bootstraps verifier and applies an exploration-enhancing reward to prevent diversity collapse; for low-confidence samples, it delegates pseudo-label selection to the verifier to filter incorrect pseudo-labels; and for medium-confidence samples, it bypasses verification entirely. Extensive experiments demonstrate that TTRL-CoCoV outperforms the best competing methods across 6 widely-recognized benchmarks, achieves average absolute gains of +9.8% in Pass@1 and +18.7% in Pass@16 over TTRL, and even achieves absolute Pass@1 improvements of up to +5.0% across multiple reasoning benchmarks when compared against fully supervised RL methods. Our code repository: https://github.com/shanjf666/CoCoV.

cs.LG↗

arXiv2026

TSQAgent: Rating Time Series Data Quality via Dedicated Agentic Reasoning

Assessing the quality of time series (TS) data is fundamental yet inherently challenging due to the multifaceted nature of quality dimensions. Recently, large language models (LLMs) have emerged as a promising paradigm for TS quality assessment via pairwise comparison and per-dimension evaluation. However, existing approaches rely on manually predefined quality dimensions and purely text-based reasoning, leaving it unknown whether LLMs can identify truly relevant quality dimensions or perform grounded and quantitative quality comparisons. To investigate this, we construct TSQBench, a dedicated benchmark for evaluating LLMs on two progressive capabilities: (i) understanding and identifying relevant quality dimensions, and (ii) performing quality comparison under specific dimensions. Our analysis reveals that current LLMs consistently struggle with both dimension identification and evidence-grounded quality comparison. To address these limitations, we propose TSQAgent, a novel agentic reasoning framework for TS quality rating consisting of three collaborative roles: Perceiver for focused dimension selection, Inspector for dimension-wise quantitative analysis, and Adjudicator that aggregates and refines the final judgment. In particular, we introduce an agentic reasoning strategy that instills the ability to identify and prioritize the most relevant quality dimensions, and further propose an agent workflow equipped with external analytical tools to enable precise quantitative comparisons over selected dimensions. Experiments on both the proposed benchmark and eleven real-world datasets demonstrate that our framework not only substantially improves LLMs' capabilities in quality understanding and quantitative comparison but also effectively translates these improvements into better quality-aware data selection, leading to enhanced downstream performance and data efficiency.

cs.AI↗

arXiv2026

Enhanced All-Distance Equi-Zenith Angle Method for Cosmic-Ray Anisotropy Measurement

Long-term observations indicate that the relative intensity of cosmic-ray anisotropy remains below $0.1\%$ for energies less than $\sim 1$ PeV. Measuring such faint signals poses a significant challenge in data analysis, requiring careful removal of instrumental and atmospheric artifacts. The all-distance equi-zenith angle method is widely employed to extract cosmic-ray anisotropies, as it effectively suppresses the instantaneous variations arising from the instrument and atmosphere. \textcolor{black}{However, instability in the detector efficiency makes precise measurements of anisotropy challenging with this method.} In this work, we present an enhanced all-distance equi-zenith angle method for cosmic-ray anisotropy measurement. Unlike previous implementations, our improved approach enables the simultaneous measurement of anisotropies over multiple time frames and allows the detection efficiency to be determined directly from the data. This feature makes the method especially suitable for applications where the detector array does not operate with long-term stability\textcolor{black}{, and thus allows for the measurement of anisotropy with high-precision}. Moreover, our enhanced method is also feasible when the data do not span complete tropical years.

astro-ph.HE↗

arXiv2026

DA-UCT: Self-Supervised Domain-Adaptive Ultrasound Computed Tomography for Rapid Musculoskeletal Sound Speed Reconstruction

Ultrasound computed tomography (UCT) via full waveform inversion (FWI) enables high-resolution quantitative imaging for tissue characterization and disease diagnosis. However, UCT suffers from large computational burden and severe convergence issues due to highly nonlinear optimization. Deep learning can accelerate UCT reconstruction, but supervised training requires large-scale labeled datasets difficult to obtain in vivo. To address these limitations, we propose SDA-UCT, a two-stage self-supervised domain-adaptive framework for rapid and accurate UCT imaging of musculoskeletal tissues. SDA-UCT employs an attention-enhanced network (AttUCT) pre-trained on simulation datasets and transfers to in-vivo data via physics-informed self-supervised learning, effectively bridging the simulation-to-real domain gap. A Low-Rank Adaptation (LoRA) mechanism is integrated to enable efficient adaptation across diverse clinical scenarios. Results showed that AttUCT achieved high-quality SOS reconstruction for simulated human forearm with a PSNR of 29.23 dB and SSIM of 0.928, outperforming conventional FWI and existing deep learning methods. Validated on in-vivo data, SDA-UCT successfully reconstructed SOS images revealing complex anatomical structures (skin, fat, muscle, tendon, bone and bone marrow) for human forearm, in high concordance with MRI references. The LoRA mechanism adjusting only 3% of parameters achieved comparable performance to full fine-tuning. The rapid reconstruction (5 ms per frame) enables real-time 3D visualization, achieving five-orders-of-magnitude improvement over traditional FWI. This work represents the first self-supervised domain-adaptive deep learning for rapid, high-resolution in-vivo UCT imaging, showing potential for musculoskeletal disease diagnosis.

cs.CV↗