Search arXiv⌕ Search

arXiv subjects

Jialin Li

Publications and source records attributed to Jialin Li.

At least 55 records · Page 3Linked to original sources
arXiv2026

SkillSmith: Compiling Agent Skills into Boundary-Guided Runtime Interfaces

Recently, skills have been widely adopted in large language model (LLM)-based agent systems across various domains. In existing frameworks, skills are typically injected into the agent reasoning loop as contextual guidance once matched to a runtime task, enabling specialized task-solving capabilities. We find that this execution paradigm introduces two major sources of redundancy: irrelevant context injection and repeated skill-specific reasoning and planning. To this end, we propose SkillSmith, a boundary-first compiler-runtime framework that compiles skill packages offline into minimal executable interfaces. By extracting fine-grained operational boundaries from skills, SkillSmith enables agents to dynamically access and execute only the relevant components at runtime, thereby minimizing unnecessary context injection and redundant reasoning overhead. In the evaluation on SkillsBench benchmark, SkillSmith reduces solve-stage token usage by 57.44%, thinking iterations by 42.99%, solve time by 50.57% (2.02x faster), and token-proportional monetary cost by 57.44% compared with using raw-skills. Moreover, compiled artifacts produced by a stronger model can be reused by a smaller or more efficient runtime model, improving task accuracy in cases where raw skill interpretation fails. The source code and data are available at https://github.com/AetherHeart-AI/Aeloon.

cs.AI↗
arXiv2026

GEAR: GEometry-motion Alternating Refinement for Articulated Object Modeling with Gaussian Splatting

High-fidelity interactive digital assets are essential for embodied intelligence and robotic interaction, yet articulated objects remain challenging to reconstruct due to their complex structures and coupled geometry-motion relationships. Existing methods suffer from instability in geometry-motion joint optimization, while their generalization remains limited on complex multi-joint or out-of-distribution objects. To address these challenges, we propose GEAR, an EM-style alternating optimization framework that jointly models geometry and motion as interdependent components within a Gaussian Splatting representation. GEAR treats part segmentation as a latent variable and joint motion parameters as explicit variables, alternately refining them for improved convergence and geometric-motion consistency. To enhance part segmentation quality without sacrificing generalization, we leverage a vanilla 2D segmentation model to provide multi-view part priors, and employ a weakly supervised constraint to regularize the latent variable. Experiments on multiple benchmarks and our newly constructed dataset GEAR-Multi demonstrate that GEAR achieves state-of-the-art results in geometric reconstruction and motion parameters estimation, particularly on complex articulated objects with multiple movable parts.

cs.CV↗
arXiv2026

PET-DINO: Unifying Visual Cues into Grounding DINO with Prompt-Enriched Training

Open-Set Object Detection (OSOD) enables recognition of novel categories beyond fixed classes but faces challenges in aligning text representations with complex visual concepts and the scarcity of image-text pairs for rare categories. This results in suboptimal performance in specialized domains or with complex objects. Recent visual-prompted methods partially address these issues but often involve complex multi-modal designs and multi-stage optimizations, prolonging the development cycle. Additionally, effective training strategies for data-driven OSOD models remain largely unexplored. To address these challenges, we propose PET-DINO, a universal detector supporting both text and visual prompts. Our Alignment-Friendly Visual Prompt Generation (AFVPG) module builds upon an advanced text-prompted detector, addressing the limitations of text representation guidance and reducing the development cycle. We introduce two prompt-enriched training strategies: Intra-Batch Parallel Prompting (IBP) at the iteration level and Dynamic Memory-Driven Prompting (DMD) at the overall training level. These strategies enable simultaneous modeling of multiple prompt routes, facilitating parallel alignment with diverse real-world usage scenarios. Comprehensive experiments demonstrate that PET-DINO exhibits competitive zero-shot object detection capabilities across various prompt-based detection protocols. These strengths can be attributed to inheritance-based philosophy and prompt-enriched training strategies, which play a critical role in building an effective generic object detector. Project page: https://fuweifuvtoo.github.io/pet-dino.

cs.CV↗
arXiv2026

Vault: Decentralized Storage Made Durable

Decentralized storage networks (DSNs) are storage systems powered by permissionless nodes. Data placement in DSNs must tolerate not only storage-device failures but also adversarial behavior that targets data availability. Byzantine nodes introduce unique challenges due to collusion and adaptive attacks. They can target specific data blocks by clustering within a block's placement group, reducing the number of rational nodes and weakening failure tolerance. In this work, we propose a global defense against Byzantine nodes across all placement groups. We introduce a node-centric approach that guarantees stable incentives for rational nodes regardless of the number of Byzantine nodes in their placement groups. Building on this approach, we design Vault, a DSN that uses sampling-based data placement with verifiable randomness. Compared with prior DSNs, this placement strategy allows Vault to scale simultaneously in storage volume, on-chain footprint, and Byzantine tolerance. Our preliminary results show that Vault achieves the desired availability with scalable storage overhead while maintaining scalable fault tolerance.

cs.DC↗
arXiv2026

Supermassive Blockchain

Storage scalability is paramount in the era of big data blockchain. A storage-scalable blockchain can effectively scale out state storage to an arbitrary number of nodes and reduce the storage pressure on each, similar to distributed databases. Prior research has extensively utilized sharding techniques to attain storage scalability; however, these approaches invariably compromise safety and liveness guarantees. In this work, we propose a novel state-execution decoupled architecture, and Supermassive Blockchain, a novel storage-scalable Byzantine fault tolerance (BFT) protocol that can sustain the deterministic security properties of conventional BFT protocols. The state management system employs erasure coding to ensure state availability with scalable storage consumption, while the global consensus and execution layers maintain robust security characteristics. Our evaluation indicates that Supermassive Blockchain achieves better storage scalability compared to prior approaches while incurring low network overhead.

cs.DC↗
arXiv2026

Enhanced Structured Lasso Pruning with Class-wise Information

Modern applications require lightweight neural network models. Most existing neural network pruning methods focus on removing unimportant filters; however, these may result in the loss of statistical information after pruning due to failing to consider the class-wise information. In this paper, we employ the structured lasso from the perspective of utilizing precise class-wise information for model pruning with the help of Information Bottleneck theory, which guides us to ensure the retention of statistical information before and after pruning. With these techniques, we propose two novel adaptive network pruning schemes in parallel: sparse graph-structured lasso pruning with Information Bottleneck (sGLP-IB) and sparse tree-guided lasso pruning with Information Bottleneck (sTLP-IB). The key component is that we prune the model filters utilizing sGLP-IB and sTLP-IB with more precise structured class-wise relatedness. Compared to multiple state-of-the-art methods, our approaches achieve the best performance across three datasets and six model structures on extensive experiments. For example, with the VGG16 model based on the CIFAR-10 dataset, we can reduce the parameters by 85%, decrease the FLOPs by 61%, and maintain an accuracy of 94.10% (0.14% better than the original). For large-scale ImageNet, we can reduce the parameters by 55% while keeping the accuracy at 76.12% (only drop 0.03%) using the ResNet architecture. In summary, we succeed in reducing the model size and computational resource usage while maintaining the effectiveness of accuracy.

cs.CV↗
arXiv2026

ClarEval: A Benchmark for Evaluating Clarification Skills of Code Agents under Ambiguous Instructions

To integrate seamlessly into real-world software engineering, Code Agents must evolve from passive instruction followers into proactive collaborative partners. However, current evaluation paradigms predominantly reward "guessing" user intent under ideal conditions, neglecting the agent's ability to align with users through dialogue--a critical trait for collaborative intelligence. In this work, we propose a paradigm shift in evaluation to drive this transition. We introduce ClarEval, a framework designed to assess an agent's "Collaborative Quotient" by simulating the inherent ambiguity of human communication. By systematically injecting three types of realistic ambiguity (missing goals, premises, and ambiguous terminology) into standard tasks, we force agents to step out of their "generator" role and engage in requirement elicitation. To quantify this capability, we propose a metric suite led by Average Turns to Clarify (ATC) and Key Question Coverage (KQC), which measure not just the correctness of the generated code, but the efficiency and precision of the collaboration. Our experiments on eleven state-of-the-art agents reveal a stark reality: while models like GPT-5-Coder excel at coding, they often lack the strategic communication skills required for efficient partnership. ClarEval thus serves as a crucial roadmap for bridging the gap between strong coders and capable collaborators.The code is available at https://github.com/JialinLi13/ClarEval

cs.SE↗
arXiv2026

Phase-Induced Amplitude Apodization Complex Mask Coronagraph (PIAACMC) on-sky demonstration with MagAO-X

Advancing the technological development of small inner working angle (IWA) coronagraphs is essential to enabling high-contrast imaging of temperate exoplanets with future extremely large telescopes. The PIAACMC has been shown to closely approach the theoretical limit for coronagraphic throughput but its performance has not been fully characterised on-sky. This study serves as the first on-sky characterisation of contrast and IWA performance of the PIAACMC and its first technological demonstration at sub-micron wavelengths. We designed and manufactured phase-shifting focal plane masks optimised for two cases, a narrowband 875 filter (875nm, 3% band) and a broadband z' filter (908nm, 14% band). We tested the coronagraphs both with an internal source and on-sky using MagAOX, the extreme adaptive optics instrument for the Magellan Clay 6.5 m telescope at Las Campanas Observatory. We show good recovery of the off-axis light's PSF shape within 92% and 97% depending on the separation when aligning the inverse set of PIAA lenses. We demonstrate sub-lambda/D IWAs of about 0.74 lambda/D in 875 and 0.76 lambda/D in z'. We reach average raw contrasts within 1 and 5 lambda/D with the internal source of about 1.6e-3 in 875 and 1.3e-3 in z'. These are mainly limited by the focal plane mask manufacturing errors, jitter, and residual quasi-static speckles in MagAO-X. We also show on-sky average raw contrasts within 1 and 5 lambda/D of about 1.4e-2 in 875 and 7.8e-3 in z'. These are likely limited by wavefront control, low-order aberrations, and poor observing conditions. Future work will improve the design and manufacturing processes of the focal plane masks to improve robustness and reach deeper contrast, as well as integrate focal plane wavefront control for non-common path aberrations correction.

astro-ph.IM↗
arXiv2026

PrefIx: Understand and Adapt to User Preference in Human-Agent Interaction

LLM-based agents can complete tasks correctly yet still frustrate users through poor interaction patterns, such as excessive confirmations, opaque reasoning, or misaligned pacing. Current benchmarks evaluate task accuracy but overlook how agents interact: whether they infer preferences from implicit cues, adapt dynamically, or maintain fine-grained interaction quality. We introduce Prefix, a configurable environment that evaluates both what agents accomplish and how they interact. Central to Prefix is the Interaction-as-a-Tool (IaaT) paradigm, which treats interaction behaviors as structured tool calls, unifying them with existing evaluation frameworks. We define 31 preference settings across 14 attributes and formalize user experience (UX) as a core metric alongside task accuracy. A composite LLM-as-a-Judge mechanism across seven UX dimensions achieves strong aggregate reliability (ICC > 0.79), high internal consistency (alpha = 0.943), and human correlation (rho = 0.52-0.78). Preference-aware agents show 7.6% average UX improvement and 18.5% gain in preference alignment. Our work is openly accessible.

cs.HC↗
arXiv2026

Cortex: Achieving Low-Latency, Cost-Efficient Remote Data Access For LLM via Semantic-Aware Knowledge Caching

Large Language Model (LLM) agents tackle data-intensive tasks such as deep research and code generation. However, their effectiveness depends on frequent interactions with knowledge sources across remote clouds or regions. Such interactions can create non-trivial latency and cost bottlenecks. Existing caching solutions focus on exact-match queries, limiting their effectiveness for semantic knowledge reuse. To address this challenge, we introduce Cortex, a novel cross-region knowledge caching architecture for LLM agents. At its core are two abstractions: Semantic Element (SE) and Semantic Retrieval Index (Seri). A semantic element captures the semantic embedding representation of an LLM query together with performance-aware metadata such as latency, cost, and staticity. Seri then provides two-stage retrieval: a vector similar index with semantic embedding for fast candidate selection and a lightweight LLM-powered semantic judger for precise validation. Atop these primitives, Cortex builds a new cache interface that includes a new semantic-aware cache hit definition, a cost-efficient eviction policy, and proactive prefetching. To reduce overhead, Cortex co-locates the small LLM judger with the main LLM using adaptive scheduling and resource sharing. Our evaluation demonstrates that Cortex delivers substantial performance improvements without compromising correctness. On representative search workloads, Cortex achieves up to a 3.6x increase in throughput by maintaining cache hit rates of over 85%, while preserving accuracy virtually identical to non-cached baselines. Cortex also improves throughput for coding tasks by 20%, showcasing its versatility across diverse agentic workloads.

cs.DC↗
arXiv2026

Revisiting Parameter Server in LLM Post-Training

Modern data parallel (DP) training favors collective communication over parameter servers (PS) for its simplicity and efficiency under balanced workloads. However, the balanced workload assumption no longer holds in large language model (LLM) post-training due to the high variance in sequence lengths. Under imbalanced workloads, collective communication creates synchronization barriers, leading to under-utilization of devices with smaller workloads. This change in training dynamics calls for a revisit of the PS paradigm for its robustness to such imbalance. We propose \textbf{On-Demand Communication (ODC)}, which adapts PS into Fully Sharded Data Parallel (FSDP) by replacing collective all-gather and reduce-scatter with direct point-to-point communication. Compared to FSDP, ODC reduces the synchronization barrier from once per layer to once per minibatch and decouples the workload on each device so that faster workers are not stalled. It also enables simpler and more effective load balancing at the minibatch level. Across diverse LLM post-training tasks, ODC consistently improves device utilization and training throughput, achieving up to a 36\% speedup over standard FSDP. These results demonstrate that ODC is a superior fit for the prevalent imbalanced workloads in LLM post-training. Our implementation of ODC and integration with FSDP is open-sourced at https://github.com/sail-sg/odc.

cs.DC↗
arXiv2026

Optimisation of the vertex detector and measurement of Higgs decays to second-generation quarks at the CEPC

The vertex detector is crucial for precision measurements of the Higgs boson at the electron-positron Higgs factory. Benchmarked with $H \to c\bar{c}$ and $H \to s\bar{s}$ measurements in the $ν\barνH$ channel, we perform an optimisation study on the inner radius and spatial resolution of the vertex detector using the Jet Origin Identification (JOI) framework, which determines the parton flavor of jets using advanced Artificial Intelligence (AI) algorithm. We observe that, compared to the reference detector configuration, halving the inner radius and spatial resolution improves the transverse and longitudinal impact parameter resolution approximately by a factor of two, while increasing the accuracy and significance of the $H \to c\bar{c}/s\bar{s}$ measurement by 4\% and 8\%, respectively. Conversely, doubling these parameters results in comparable degradation, with variations in the inner radius being the dominant factor. Our results provide guidance for detector design and highlight promising prospects for identifying the $H \to s\bar{s}$ decay mode at future Higgs factories.

hep-ex↗
arXiv2026

Beyond One-Size-Fits-All: A Survey of Personalized Affective Computing in Human-Agent Interaction

In personalized machine learning, the aim of personalization is to train a model that caters to a specific individual or group of individuals by optimizing one or more performance metrics and adhering to specific constraints. In this paper, we discuss the need for personalization in affective computing and present the first survey of existing approaches for personalization in affective computing. Our review spans training techniques and objectives towards the personalization of affective computing models across various interaction modes and contexts. We develop a taxonomy that clusters existing approaches into Data-level and Model-level approaches. Across the Data-Level and Model-Level broad categories, we group existing approaches into seven sub-categories: (1) User-Specific Models, (2) Group-Specific Models, (3) Weighting-Based Approaches, (4) Feature Augmentation, (5) Generative-Based Models which fall into the Data-Level approaches, (6) Fine-Tuning Approaches, and (7) Multitask Learning Approaches falling under the model-level approaches. We provide a problem formulation for personalized affective computing, and to each of the identified sub-categories. Additionally, we provide a statistical analysis of the surveyed literature, analyzing the prevalence of different affective computing tasks, interaction modes (i.e. Human-Computer Interaction (HCI), Human-Human interaction (HHI), Human-Robot Interaction (HRI)), interaction contexts (e.g. educative, social, gaming, etc.), and the level of personalization among the surveyed works. Based on our analysis, we provide a road-map for researchers interested in exploring this direction.

cs.HC↗
arXiv2026

DeepH-pack: A general-purpose neural network package for deep-learning electronic structure calculations

In computational physics and materials science, first-principles methods, particularly density functional theory, have become central tools for electronic structure prediction and materials design. Recently, rapid advances in artificial intelligence (AI) have begun to reshape the research landscape, giving rise to the emerging field of deep-learning electronic structure calculations. Despite numerous pioneering studies, the field remains in its early stages; existing software implementations are often fragmented, lacking unified frameworks and standardized interfaces required for broad community adoption. Here we present DeepH-pack, a comprehensive and unified software package that integrates first-principles calculations with deep learning. By incorporating fundamental physical principles into neural-network design, such as the nearsightedness principle and the equivariance principle, DeepH-pack achieves robust cross-scale and cross-material generalizability. This allows models trained on small-scale structures to generalize to large-scale and previously unseen materials. The toolkit preserves first-principles accuracy while accelerating electronic structure calculations by several orders of magnitude, establishing an efficient and intelligent computational paradigm for large-scale materials simulation, high-throughput materials database construction, and AI-driven materials discovery.

cond-mat.mtrl-sci↗
arXiv2025

Reaching Agreement Among Reasoning LLM Agents

Multi-agent systems have extended the capability of agentic AI. Instead of single inference passes, multiple agents perform collective reasoning to derive high quality answers. However, existing multi-agent orchestration relies on static heuristic workflows such as fixed loop limits and barrier synchronization. These ad-hoc approaches waste computational resources, incur high latency due to stragglers, and risk finalizing transient agreements. We argue that reliable multi-agent reasoning requires a formal foundation analogous to classical distributed consensus problem. To that end, we propose a formal model of the multi-agent refinement problem. The model includes definitions of the correctness guarantees and formal semantics of agent reasoning. We then introduce Aegean, a consensus protocol designed for stochastic reasoning agents that solves multi-agent refinement. We implement the protocol in Aegean-Serve, a consensus-aware serving engine that performs incremental quorum detection across concurrent agent executions, enabling early termination when sufficient agents converge. Evaluation using four mathematical reasoning benchmarks shows that Aegean provides provable safety and liveness guarantees while reducing latency by 1.2--20$\times$ compared to state-of-the-art baselines, maintaining answer quality within 2.5%. Consistent gains across both local GPU deployments and commercial API providers validate that consensus-based orchestration eliminates straggler delays without sacrificing correctness.

cs.DC↗
arXiv2025

UniFS: Unified Multi-Contrast MRI Reconstruction via Frequency-Spatial Fusion

Recently, Multi-Contrast MR Reconstruction (MCMR) has emerged as a hot research topic that leverages high-quality auxiliary modalities to reconstruct undersampled target modalities of interest. However, existing methods often struggle to generalize across different k-space undersampling patterns, requiring the training of a separate model for each specific pattern, which limits their practical applicability. To address this challenge, we propose UniFS, a Unified Frequency-Spatial Fusion model designed to handle multiple k-space undersampling patterns for MCMR tasks without any need for retraining. UniFS integrates three key modules: a Cross-Modal Frequency Fusion module, an Adaptive Mask-Based Prompt Learning module, and a Dual-Branch Complementary Refinement module. These modules work together to extract domain-invariant features from diverse k-space undersampling patterns while dynamically adapt to their own variations. Another limitation of existing MCMR methods is their tendency to focus solely on spatial information while neglect frequency characteristics, or extract only shallow frequency features, thus failing to fully leverage complementary cross-modal frequency information. To relieve this issue, UniFS introduces an adaptive prompt-guided frequency fusion module for k-space learning, significantly enhancing the model's generalization performance. We evaluate our model on the BraTS and HCP datasets with various k-space undersampling patterns and acceleration factors, including previously unseen patterns, to comprehensively assess UniFS's generalizability. Experimental results across multiple scenarios demonstrate that UniFS achieves state-of-the-art performance. Our code is available at https://github.com/LIKP0/UniFS.

cs.CV↗
arXiv2025

Sympathetic Cooling of Levitated Optomechanics through Nonreciprocal Coupling

Optomechanical cooling of levitated nanoparticles has become an essential topic in modern quantum physics, providing a platform for exploring macroscopic quantum phenomena and high-precision sensing. However, conventional cavity-assisted cooling is fundamentally constrained by cavity dissipation and environmental noise, limiting the attainable minimum temperature. In this work, we propose a non-Hermitian optomechanical cooling scheme through nonreciprocal coupling between two levitated nanoparticles, where one particle is directly cooled by an optical cavity and the other is cooled indirectly through a non-Hermitian interaction. Both analytical solutions and numerical simulations reveal that increasing nonreciprocity enhances directional energy transfer, enabling the target particle to reach a lower phonon occupation than is achievable in conventional cavity cooling. This study demonstrates a new cooling mechanism driven by non-Hermitian interactions, offering theoretical guidance for realizing controllable energy flow and deep cooling in levitated optomechanical systems, and paving the way for future developments in quantum control and sensing technologies.

quant-ph↗
arXiv2025

FlashMP: Fast Discrete Transform-Based Solver for Preconditioning Maxwell's Equations on GPUs

Efficiently solving large-scale linear systems is a critical challenge in electromagnetic simulations, particularly when using the Crank-Nicolson Finite-Difference Time-Domain (CN-FDTD) method. Existing iterative solvers are commonly employed to handle the resulting sparse systems but suffer from slow convergence due to the ill-conditioned nature of the double-curl operator. Approximate preconditioners, like Successive Over-Relaxation (SOR) and Incomplete LU decomposition (ILU), provide insufficient convergence, while direct solvers are impractical due to excessive memory requirements. To address this, we propose FlashMP, a novel preconditioning system that designs a subdomain exact solver based on discrete transforms. FlashMP provides an efficient GPU implementation that achieves multi-GPU scalability through domain decomposition. Evaluations on AMD MI60 GPU clusters (up to 1000 GPUs) show that FlashMP reduces iteration counts by up to 16x and achieves speedups of 2.5x to 4.9x compared to baseline implementations in state-of-the-art libraries such as Hypre. Weak scalability tests show parallel efficiencies up to 84.1%.

cs.DC↗