Searcharxiv⌕ Search

arXiv subjects

Libo Zhang

Publications and source records attributed to Libo Zhang.

At least 19 recordsLinked to original sources
arXiv2026

Role-aware Heuristic Episodic Attention for Conversational LLMs

Large language models often lose track of persistent instructions and relevant information as multi-turn conversations grow. We study this cumulative contextual decay through three related failure modes: attention pollution, dilution, and drift. We propose REA(Role-aware Heuristic Episodic Attention), a context-management framework that assigns different persistence and representation policies to instructions and episodic interactions. Instructional Memory retains identified global constraints in a dedicated prefix. Episodic Memory preserves user inputs and compresses model replies, while heuristic retrieval selects raw text, compressed representations, or omission for each historical turn. On Long-MT-Bench+, REA improves the judge score from 6.32 to 7.36 on a 10-point scale, a 16.5% relative gain over the Vanilla baseline, and reduces average latency by 2.91x. Additional evaluations show aggregate gains on three backbones spanning 1.7B-7B parameters and on Chinese and English role-playing tasks. These results support role-aware context management as a practical approach to maintaining conversational continuity and instruction adherence.

cs.CL↗
arXiv2026

Role-aware Heuristic Episodic Attention for Conversational LLMs

Large language models often lose track of persistent instructions and relevant information as multi-turn conversations grow. We study this cumulative contextual decay through three related failure modes: attention pollution, dilution, and drift. We propose REA (Role-aware Heuristic Episodic Attention), a context-management framework that assigns different persistence and representation policies to instructions and episodic interactions. Instructional Memory retains identified global constraints in a dedicated prefix. Episodic Memory preserves user inputs and compresses model replies, while heuristic retrieval selects raw text, compressed representations, or omission for each historical turn. On Long-MT-Bench+, REA improves the judge score from 6.32 to 7.36 on a 10-point scale, a 16.5% relative gain over the Vanilla baseline, and reduces average latency by 2.91$\times$. Additional evaluations show aggregate gains on three backbones spanning 1.7B-7B parameters and on Chinese and English role-playing tasks. These results support role-aware context management as a practical approach to maintaining conversational continuity and instruction adherence.

cs.CL↗
arXiv2026

PGL-3D: Towards Progressive Geometric Learning for 3D Visual Query Localization

3D Visual Query Localization (3DVQL) retrieves the latest contiguous occurrence of a queried object in an RGB--point-cloud sequence and predicts a 9-DoF cuboid for every response frame. The query is captured independently of the search sequence, so its annotated pose may differ from how the object appears in the search frames. The benchmark baseline predicts cuboids after feature modeling, leaving their geometry unused for subsequent feature refinement. We investigate whether complete intermediate cuboids can improve query and proposal representations before final decoding. We introduce Progressive Geometric Learning for 3DVQL (PGL-3D), a predict--select--refine--re-predict framework that uses intermediate cuboids to guide the aggregation of search evidence and update query and proposal representations. A shared head first predicts a complete cuboid for every proposal. Query--Tube--Memory (QTM) then selects reference observations by combining proposal association, cuboid quality, frame response, and target absence, since association confidence alone establishes neither target presence nor geometric accuracy. The center, size, and orientation of each selected cuboid define soft pooling weights over query-conditioned proposal features. The pooled memory updates the query and proposal representations, and the head re-predicts from the updated features. A training-only objective, ST-D9O, supervises cuboid geometry at every stage by adding boundary, signed-distance, and soft-overlap terms to parameter regression. PGL-3D achieves a mean stAP of $0.270 \pm 0.004$ on 3DVQL, compared with $0.044$ reported for LaF. Ablations support the benefits of geometry-guided feature updates, while stage-wise analyses show improved cuboid accuracy. Replacing the geometry objective in our PROT3D reproduction with ST-D9O improves mAO on GSOT3D from $21.63\%$ to $25.78\%$. Our code and models will be released.

cs.CV↗
arXiv2026

Three-Phase Scribble-Adaptive Curriculum Learning for autoPETV Grand Challenge

This report describes Libo Zhang's algorithmic solution to autoPETV Grand Challenge on interactive lesion segmentation in whole-body PET/CT. Interaction is encoded as two additional input channels that rasterize the accumulated foreground and background scribbles, and a residual-encoder U-Net of about 140 million parameters is trained with a three-phase curriculum over 4000 epochs: the network first learns fully automatic segmentation with silent interaction channels, then observes ground-truth-derived scribbles under randomly sampled visibility modes, and finally adapts to its own mistakes through online simulation of up to five error-driven correction steps. Training draws on 1811 autoPET and DeepPSMA studies, and the submission ensembles the best and final checkpoints of five folds by logit averaging. In interactive five-fold cross-validation with six interaction steps, the final checkpoints reach a mean AUC-Dice of 3.836 and a mean AUC-DMM of 3.869, improving monotonically in every fold, with roughly half of the total gain delivered by the first corrective scribble. Our code and trained model checkpoints are available on https://github.com/Libo1023/autoPETV-Curriculum.

cs.CV↗
arXiv2026

LoopTrack: A Simple Baseline for Parameter-Efficient Transformer Tracking

Current Transformer-based tracking methods typically stack multiple Transformer blocks with separate parameters to model interactions between the target template and the search region for target localization. These trackers often incur substantial parameter overhead from stacked blocks, making their deployment on resource-limited devices difficult. To address this, we propose a parameter-efficient Transformer tracking framework, dubbed LoopTrack, which repeatedly applies a set of Transformer blocks with shared parameters to interact features in a looped architecture for tracking, significantly reducing the number of parameters. To further exploit target cues, we present two lightweight designs, including target-aware looping (TAL) and gated target memory (GTM). The former applies intermediate target information generated by one loop to guide feature interaction in the subsequent loop, enabling progressive feature refinement, while the latter maintains a compact memory across frames, which is incorporated into the loop process to provide long-term information to the tracker, mitigating temporal drift in tracking. Compared to existing Transformer trackers, LoopTrack enables multiple rounds of feature interaction with fewer model parameters, making it resource-friendly for deployment. In extensive experiments on multiple datasets, LoopTrack shows a favorable accuracy-parameter trade-off. In particular, our LoopTrack$_{\rm One}$, with a single shared Transformer block, achieves 66.2\% SUC score on LaSOT with only 3.4M parameters, while LoopTrack$_{\rm Three}$, using three shared blocks, achieves 69.3\% SUC score with 6.4M parameters, surpassing existing parameter-efficient tracking methods with comparable or larger model size. With LoopTrack, we aim to establish a simple yet strong baseline for parameter-efficient Transformer tracking. Our code and models will be released.

cs.CV↗
arXiv2026

Exploiting Target Knowledge from MLLMs for Robust Few-Shot Segmentation

Few-shot segmentation (FSS) aims to segment unseen object categories with a few (e.g., one or five) labeled examples, enabling efficient adaptation to novel classes. Conventional models typically rely on appearance-based visual matching between support and query images for segmentation. While straightforward, these methods often struggle to handle significant appearance discrepancies and occlusions in the query image due to insufficient target knowledge. To mitigate this, we introduce a novel framework that mines target knowledge using the strong reasoning capacity of Multimodal Large Language Models (MLLMs) and employs it to enhance FSS. Specifically, building on SAM 2, our method, named MK-FSS, exploits two forms of complementary knowledge derived from a query image by an MLLM for FSS, including spatial knowledge, which provides a spatial prior indicating the potential target location, and semantic knowledge, which describes the target using text. The spatial knowledge is first encoded into a memory representation, and then resulting memory is integrated with the support-guided memory feature from query image through a carefully designed dual-memory debate-fusion (DMDF) module, yielding a more robust target memory feature. In parallel, the semantic knowledge is encoded into the textual feature, which is fused with multi-scale query features via a progressive cross-modal prompt generator (PCPG), producing a target-aware multimodal prompt for segmentation. Working together, the dual-memory feature and the multimodal prompt provide a comprehensive representation of the target, enabling more robust segmentation. In our extensive experiments, MK-FSS shows promising results and largely surpasses existing methods. Code will be released.

cs.CV↗
arXiv2026

Demonstrating advantages of dynamic quantum circuits on a hybrid superconducting qubit-cavity processor

Dynamic quantum circuits (DQCs) provide a hardware-efficient route to quantum computing by reducing physical-qubit overhead and compressing circuit topology through mid-circuit measurements, qubit reset and reuse, and classical feed-forward control. Here, we demonstrate the advantages of DQCs on a single hybrid superconducting qubit-cavity processor by implementing a hierarchy of algorithms with increasing complexity. This hybrid architecture consists of a high-dimensional cavity qudit serving as the computational register and a dispersively coupled superconducting transmon ancilla that is repeatedly measured, reset, and reused to enable dynamic control. Using this device, we implement a 10-bit Bernstein-Vazirani algorithm with an average success probability of 82%, surpassing state-of-the-art dynamic and static implementations in both scale and performance; an 8-bit quantum phase-estimation protocol with estimation errors below 10-3; and the first dynamic-circuit implementation of Shor's algorithm on a superconducting platform, factoring 15 over all coprime bases with squared statistical overlap values above 99.8%. These results provide concrete benchmarks for future DQC implementations and highlight the versatile advantages of DQCs with the hybrid qubit-qudit architecture, establishing it as a promising route toward scalable, programmable quantum computation.

quant-ph↗
arXiv2026

Weakly Non-Negative Supermartingales for Omega-Regular Verification

Martingale-based methods are central to probabilistic program verification, but strong global non-negativity requirements can exclude simple certificates from tractable template classes. Relaxing this requirement enlarges the search space for automated synthesis, but naive relaxations are unsound in the probabilistic setting. We introduce lazy Streett supermartingales and their lexicographic extension, showing that weak non-negativity can nevertheless be used soundly to certify almost-sure satisfaction of $ω$-regular properties with polynomial templates under a broad class of sampling distributions, including all bounded-support distributions. This extends prior weakly non-negative methods from termination to general $ω$-regular verification. We further give a compositional account of lexicographic certificates in terms of one-dimensional ones. Experiments on 170 polynomial probabilistic-program benchmarks show increases of 20.0-23.5 percentage points in verification success over the strongly non-negative baseline.

cs.PL↗
arXiv2026

Towards Long-Form Spatio-Temporal Video Grounding

In real scenarios, videos can span several minutes or even hours. However, existing research on spatio-temporal video grounding (STVG), given a textual query, mainly focuses on localizing targets in short videos of tens of seconds, typically less than one minute, which limits real-world applications. In this paper, we explore Long-Form STVG (LF-STVG), which aims to locate targets in long-term videos. Compared with short videos, long-term videos contain much longer temporal spans and more irrelevant information, making it difficult for existing STVG methods that process all frames at once. To address this challenge, we propose an AutoRegressive Transformer architecture for LF-STVG, termed ART-STVG. Unlike conventional STVG methods that require the entire video sequence to make predictions at once, ART-STVG treats the video as streaming input and processes frames sequentially, enabling efficient handling of long videos. To model spatio-temporal context, we design spatial and temporal memory banks and apply them to the decoders. Since memories from different moments are not always relevant to the current frame, we introduce simple yet effective memory selection strategies to provide more relevant information to the decoders, significantly improving performance. Furthermore, instead of parallel spatial and temporal localization, we propose a cascaded spatio-temporal design that connects the spatial decoder to the temporal decoder, allowing fine-grained spatial cues to assist complex temporal localization in long videos. Experiments on newly extended LF-STVG datasets show that ART-STVG significantly outperforms state-of-the-art methods, while achieving competitive performance on conventional short-form STVG. Our code is at: https://github.com/HengLan/ART-STVG.

cs.CV↗
arXiv2026

Nightjar: Dynamic Adaptive Speculative Decoding for Large Language Models Serving

Speculative decoding (SD) accelerates LLM inference by verifying draft tokens in parallel. However, this method presents a critical trade-off: it improves throughput in low-load, memory-bound systems but degrades performance in high-load, compute-bound environments due to verification overhead. Existing speculative decoding methods use fixed lengths and cannot adapt to workload changes or decide when to stop speculation. The cost of restarting speculative inference also remains unquantified. Under high load, the benefit of speculation diminishes, while retaining the draft model reduces KV cache capacity, limiting batch size and degrading throughput. To overcome this, we propose Nightjar, a resource-aware adaptive speculative framework. It first adjusts to the request load by dynamically selecting the optimal speculative length for different batch sizes. Crucially, Nightjar proactively disables speculative decoding when the MAB planner determines that speculation is no longer beneficial, and during the disabled phase, offloads the draft model to the CPU only under GPU memory pressure. This reclaims memory for the KV cache, thereby facilitating larger batch sizes and maximizing overall system throughput. Experiments show that Nightjar achieves up to 14.76% higher throughput than standard speculative decoding and up to 20.18% lower latency in the main benchmark suite under dynamic request arrival rates for real-time LLM serving scenarios.

cs.DC↗
arXiv2026

TuneAgent: Agentic Operating System Kernel Tuning with Reinforcement Learning

Linux kernel tuning is essential for optimizing operating system (OS) performance, yet remains challenging due to the complex kernel space, sparse performance feedback, and strong workload sensitivity. We present TuneAgent, an agentic Linux kernel tuning framework powered by rule-based reinforcement learning (RL). TuneAgent formulates the kernel space as a constrained RL environment, enabling large language models (LLMs) to autonomously explore the kernel while enforcing valid and precise configuration modifications. To address sparse performance feedback, we design structured reward functions that jointly promote reasoning standardization, configuration correctness, and performance awareness. Furthermore, we propose a two-phase training strategy that first ensures format and semantic correctness and then transitions to performance-driven exploration, accelerating convergence and reducing overhead. Experimental results show that TuneAgent consistently outperforms existing baselines, achieving up to 5.6% relative overall performance improvement while maintaining high configuration validity. We further demonstrate its robustness across multiple real-world applications, highlighting its practicality and adaptability in diverse deployment environments.

cs.LG↗
arXiv2026

Quantum-Enhanced Dark Matter Search Using Cat States

Quantum metrology has recently emerged as a powerful approach for dark matter (DM) searches, particularly using nonclassical bosonic states in microwave cavities that are sensitive to weak signals. Nonclassical cat states - macroscopic superpositions of coherent states featuring sub-Planck interference structures - offer promising advantages for high-precision measurements. However, their practical utility in DM search remains unexplored. Here, we report the first experimental application of four-component cat states within a high-quality superconducting microwave cavity to search for dark photons, a potential DM candidate. We demonstrate an 8.1-fold enhancement in the signal photon rate and constrain the dark photon kinetic mixing angle to an unprecedented $ε< 7.32 \times 10^{-16}$ near 6.44~GHz (26.6~$μ$eV). By employing a parametric sideband drive to actively tune the cavity frequency, we achieve dark photon searches and background subtraction across multiple frequency bins, yielding a sensitivity at the $10^{-16}$ level within a 100~kHz bandwidth. Our cat-assisted DM (CaD) search and frequency-scanning techniques demonstrate substantial improvements over previous results, promising potential implications in quantum-enhanced searches for new physics.

quant-ph↗
arXiv2026

Towards Visual Query Localization in the 3D World

Visual query localization (VQL) aims to predict the spatio-temporal response of the most recent occurrence in a sequence given a query. Currently, most research focuses on visual query localization in 2D videos, while its counterpart in 3D space has received little attention. In this paper, we make the first attempt to address visual query localization in the 3D world by introducing a novel benchmark, dubbed 3DVQL. Specifically, 3DVQL contains 2,002 sequences with around 170,000 frames and 6.4K response track segments from 38 object categories. Each sequence in 3DVQL is provided with multiple modalities, including point clouds, RGB images, and depth images, to support flexible research. To ensure high-quality annotations, each sequence is manually annotated with multiple rounds of verification and refinement. To the best of our knowledge, 3DVQL is the first benchmark for 3D multimodal visual query localization. To facilitate comparison in subsequent research, we implement a series of representative 3D multimodal VQL baselines using point clouds and RGB images. The experimental results show that existing methods exhibit significant performance variations across different fusion modules. To encourage future research, we propose a lift-and-attention fusion algorithm named LaF, which significantly outperforms existing baseline models. Our benchmark and model will be publicly released at https://github.com/wuhengliangliang/3DVQL.

cs.CV↗
arXiv2026

Exploring Concept Subspace for Self-explainable Text-Attributed Graph Learning

We introduce Graph Concept Bottleneck (GCB) as a new paradigm for self-explainable text-attributed graph learning. GCB maps graphs into a subspace, concept bottleneck, where each concept is a meaningful phrase, and predictions are made based on the activation of these concepts. Unlike existing interpretable graph learning methods that primarily rely on subgraphs as explanations, the concept bottleneck provides a new form of interpretation. To refine the concept space, we apply the information bottleneck principle to focus on the most relevant concepts. This not only yields more concise and faithful explanations but also explicitly guides the model to "think" toward the correct decision. We empirically show that GCB achieves intrinsic interpretability with accuracy on par with black-box Graph Neural Networks. Moreover, it delivers better performance under distribution shifts and data perturbations, showing improved robustness and generalizability, benefitting from concept-guided prediction.

cs.LG↗
arXiv2026

Robust Ego-Exo Correspondence with Long-Term Memory

Establishing object-level correspondence between egocentric and exocentric views is essential for intelligent assistants to deliver precise and intuitive visual guidance. However, this task faces numerous challenges, including extreme viewpoint variations, occlusions, and the presence of small objects. Existing approaches usually borrow solutions from video object segmentation models, but still suffer from the aforementioned challenges. Recently, the Segment Anything Model 2 (SAM 2) has shown strong generalization capabilities and excellent performance in video object segmentation. Yet, when simply applied to the ego-exo correspondence (EEC) task, SAM 2 encounters severe difficulties due to ineffective ego-exo feature fusion and limited long-term memory capacity, especially for long videos. Addressing these problems, we propose a novel EEC framework based on SAM 2 with long-term memories by presenting a dual-memory architecture and an adaptive feature routing module inspired by Mixture-of-Experts (MoE). Compared to SAM 2, our approach features (i) a Memory-View MoE module which consists of a dual-branch routing mechanism to adaptively assign contribution weights to each expert feature along both channel and spatial dimensions, and (ii) a dual-memory bank system with a simple yet effective compression strategy to retain critical long-term information while eliminating redundancy. In the extensive experiments on the challenging EgoExo4D benchmark, our method, dubbed LM-EEC, achieves new state-of-the-art results and significantly outperforms existing methods and the SAM 2 baseline, showcasing its strong generalization across diverse scenarios. Our code and model are available at https://github.com/juneyeeHu/LM-EEC.

cs.CV↗
arXiv2026

Experimental observation of exact quantum critical states

Anderson localization physics features three fundamental types of eigenstates: extended, localized, and critical, with the third one exhibiting the exotic properties in-between the former two. Confirming the presence of critical states is challenging, as it typically necessitates either advancing the analysis to the thermodynamic limit or identifying a universal mechanism which can rigorously determine these states. Here we report the unambiguous experimental realization of critical states, governed by a rigorous mechanism for exact quantum critical states, and further observe a generalized mechanism that quasiperiodic zeros in hopping couplings protect the critical states. We implement a programmable quasiperiodic mosaic model with tunable couplings and on-site potentials through a multiple superconducting qubit quantum system. By measuring the time-evolving observables, we identify the coexisting delocalized dynamics and incommensurately distributed zeros in the couplings, which are the defining features of the critical states. We map the localized-to-critical phase transition and demonstrate that critical states persist until quasiperiodic zeros are removed by strong long-range couplings, highlighting a novel generalized mechanism discovered in this experiment and shown with rigorous theory. Finally, we resolve the energy-dependent transition between localized and critical states, revealing the presence of anomalous mobility edges.

quant-ph↗
arXiv2026

Collective Dynamics in Circuit Quantum Acoustodynamics with a Macroscopic Resonator

Collective dynamics in engineered quantum systems offer a unique and versatile platform for exploring how many-body correlations bridge microscopic entanglement and macroscopic behavior. In this work, we report collective Dicke dynamics of acoustic modes in a macroscopic high-overtone bulk acoustic resonator (HBAR). To achieve this, we engineer a hybrid quantum acoustodynamic system comprising an HBAR strongly coupled to a superconducting transmon qubit. The HBAR device is distinctive in the sense that its narrow mode spacing, together with enhanced qubit-mode coupling strength, gives rise to efficient coupling between the transmon and clusters of near-resonant modes. By harnessing the system properties, we observe collective dynamics involving clusters composed by two or three mechanical modes, where their non-resonant spectrum allows for the observation of the transition between the Dicke static regime to dynamically induced timed-Dicke one. The coherent collective behavior of the system is supported by time-domain measurements of the qubit's purity, indicating the quantum nature of the collective dynamics. Overall, our work establishes HBAR-based hybrid quantum system as a promising platform for exploring many-body collective dynamics in macroscopic mechanical systems.

quant-ph↗
arXiv2026

DMTrack: Spatio-Temporal Multimodal Tracking via Dual-Adapter

In this paper, we explore adapter tuning and introduce a novel dual-adapter architecture for spatio-temporal multimodal tracking, dubbed DMTrack. The key of our DMTrack lies in two simple yet effective modules, including a spatio-temporal modality adapter (STMA) and a progressive modality complementary adapter (PMCA) module. The former, applied to each modality alone, aims to adjust spatio-temporal features extracted from a frozen backbone by self-prompting, which to some extent can bridge the gap between different modalities and thus allows better cross-modality fusion. The latter seeks to facilitate cross-modality prompting progressively with two specially designed pixel-wise shallow and deep adapters. The shallow adapter employs shared parameters between the two modalities, aiming to bridge the information flow between the two modality branches, thereby laying the foundation for following modality fusion, while the deep adapter modulates the preliminarily fused information flow with pixel-wise inner-modal attention and further generates modality-aware prompts through pixel-wise inter-modal attention. With such designs, DMTrack achieves promising spatio-temporal multimodal tracking performance with merely 0.93M trainable parameters. Extensive experiments on five benchmarks demonstrate that DMTrack achieves state-of-the-art results. Our code and models will be available at https://github.com/Nightwatch-Fox11/DMTrack.

cs.CV↗