BZPEERPapers, people, evidence and research paths
Menu

Discover

SearchFind papers, people, topics, institutions and opportunities from one place.GraphSee how papers, authors, topics and evidence connect.

Research tools

HomeContinue from the papers, people and topics that matter now.CollectionsTurn saved papers into reading lists, evidence maps and shareable context.ResearchReconstruct the path that led you from search to paper to question.CollaborationMove from a paper, topic or expert request into a focused thread.PublishCreate full-text research articles with metadata, formulas and versions.

Network

ExpertsFind specialists whose work and reviews explain why they can help.CommunitiesJoin research groups organized around papers, topics and shared evidence.

Opportunities

ListingsBrowse roles, calls and collaborations connected to real research context.

Account

Log inReturn to your saved papers, graph views and active threads.Create accountStart saving papers, building a research trail and joining teams.
Log inCreate account

Source author record

Shiyu Zhao

Shiyu Zhao appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record
Systems and Controlmath.OCArtificial IntelligenceComputer VisionMachine LearningRobotics

Catalog footprint

What is connected

16works
6topics
4close collaborators

Actions

Connect this record

Log in to claim
Open graphBrowse works

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

16 published item(s)

preprint2026arXiv

DARE: Difficulty-Adaptive Reinforcement Learning with Co-Evolved Difficulty Estimation

Reinforcement learning improves the reasoning ability of large language models but remains costly and sample-inefficient, as many rollouts provide weak learning signals. Difficulty-aware data selection methods attempt to address this by prioritizing moderately difficult prompts, yet our analysis reveals three limitations: difficulty estimates become inaccurate under policy drift, data selection alone yields limited final-performance gains, and inference efficiency remains largely unchanged. These findings suggest that efficient and effective RL requires more than filtering by difficulty: the policy should learn to solve hard tasks while producing concise responses for easy ones. To this end, we propose **Dare**, a unified framework that co-evolves difficulty estimation with the policy via self-normalized importance sampling, maintains diverse difficulty coverage through a symmetric Beta sampling distribution, and applies tailored training strategies across difficulty tiers with adaptive compute allocation. Extensive experiments across multiple models and domains demonstrate that **Dare** consistently outperforms existing methods in training efficiency, final effectiveness, and inference efficiency, producing more concise responses on easy tasks while improving correctness on hard ones. Code is available at https://github.com/EtaYang10th/DARE.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Evidence Over Plans: Online Trajectory Verification for Skill Distillation

Agent skills can remarkably improve task success rates by using human-written procedural documents, but their quality is difficult to assess without environment-grounded verification. Existing skill generation methods heavily rely on preference logs rather than direct environment interaction, often yielding negligible or even degraded gains. We identify that it is a fundamental timing bottleneck: robust skills should be posterior-based, distilled from empirical environment interaction rather than prior plans. In this study, we introduce the Posterior Distillation Index (PDI), a trajectory-level metric that quantifies how well a distilled skill is grounded in the task-environment evidence. To operationalize PDI, we present SPARK (Structured Pipelines for Autonomous Runnable tasKs and sKill generation) for preserving task execution evidence towards full trajectory-level analysis. SPARK generates environment-verified trajectories used to compute PDI, and it applies PDI as an online diagnostic and intervention signal to ensure posterior skill formation. Across 86 runnable tasks, SPARK-generated skills consistently surpass no-skill baselines and outperform human-written skills on student models (inference cost up to 1,000x cheaper than teacher models). These findings show that PDI-guided distillation produces efficient and transferable skills grounded in the task-environment interaction. We release our code at https://github.com/EtaYang10th/spark-skills .

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Observability-Enhanced Target Motion Estimation via Bearing-Box: Theory and MAV Applications

Monocular vision-based target motion estimation is a fundamental challenge in numerous applications. This work introduces a novel bearing-box approach that fully leverages modern 3D detection measurements that are widely available nowadays but have not been well explored for motion estimation so far. Unlike existing methods that rely on restrictive assumptions such as isotropic target shape and lateral motion, our bearing-box estimator can estimate both the target's motion and its physical size without these assumptions by exploiting the information buried in a 3D bounding box. When applied to multi-rotor micro aerial vehicles (MAVs), the estimator yields an interesting advantage: it further removes the need for higher-order motion assumptions by exploiting the unique coupling between MAV's acceleration and thrust. This is particularly significant, as higher-order motion assumptions are widely believed to be necessary in state-of-the-art bearing-based estimators. We support our claims with rigorous observability analyses and extensive experimental validation, demonstrating the estimator's superior performance in real-world scenarios.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Exploiting Unlabeled Data with Vision and Language Models for Object Detection

Building robust and generic object detection frameworks requires scaling to larger label spaces and bigger training datasets. However, it is prohibitively costly to acquire annotations for thousands of categories at a large scale. We propose a novel method that leverages the rich semantics available in recent vision and language models to localize and classify objects in unlabeled images, effectively generating pseudo labels for object detection. Starting with a generic and class-agnostic region proposal mechanism, we use vision and language models to categorize each region of an image into any object category that is required for downstream tasks. We demonstrate the value of the generated pseudo labels in two specific tasks, open-vocabulary detection, where a model needs to generalize to unseen object categories, and semi-supervised object detection, where additional unlabeled images can be used to improve the model. Our empirical evaluation shows the effectiveness of the pseudo labels in both tasks, where we outperform competitive baselines and achieve a novel state-of-the-art for open-vocabulary object detection. Our code is available at https://github.com/xiaofeng94/VL-PLM.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Global Matching with Overlapping Attention for Optical Flow Estimation

Optical flow estimation is a fundamental task in computer vision. Recent direct-regression methods using deep neural networks achieve remarkable performance improvement. However, they do not explicitly capture long-term motion correspondences and thus cannot handle large motions effectively. In this paper, inspired by the traditional matching-optimization methods where matching is introduced to handle large displacements before energy-based optimizations, we introduce a simple but effective global matching step before the direct regression and develop a learning-based matching-optimization framework, namely GMFlowNet. In GMFlowNet, global matching is efficiently calculated by applying argmax on 4D cost volumes. Additionally, to improve the matching quality, we propose patch-based overlapping attention to extract large context features. Extensive experiments demonstrate that GMFlowNet outperforms RAFT, the most popular optimization-only method, by a large margin and achieves state-of-the-art performance on standard benchmarks. Thanks to the matching and overlapping attention, GMFlowNet obtains major improvements on the predictions for textureless regions and large motions. Our code is made publicly available at https://github.com/xiaofeng94/GMFlowNet

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Mask and Reason: Pre-Training Knowledge Graph Transformers for Complex Logical Queries

Knowledge graph (KG) embeddings have been a mainstream approach for reasoning over incomplete KGs. However, limited by their inherently shallow and static architectures, they can hardly deal with the rising focus on complex logical queries, which comprise logical operators, imputed edges, multiple source entities, and unknown intermediate entities. In this work, we present the Knowledge Graph Transformer (kgTransformer) with masked pre-training and fine-tuning strategies. We design a KG triple transformation method to enable Transformer to handle KGs, which is further strengthened by the Mixture-of-Experts (MoE) sparse activation. We then formulate the complex logical queries as masked prediction and introduce a two-stage masked pre-training strategy to improve transferability and generalizability. Extensive experiments on two benchmarks demonstrate that kgTransformer can consistently outperform both KG embedding-based baselines and advanced encoders on nine in-domain and out-of-domain reasoning tasks. Additionally, kgTransformer can reason with explainability via providing the full reasoning paths to interpret given answers.

0 citations0 reviewsNo code linkedOpen access
preprint2016arXiv

Localizability and Distributed Protocols for Bearing-Based Network Localization in Arbitrary Dimensions

This paper addresses the problem of bearing-based network localization, which aims to localize all the nodes in a static network given the locations of a subset of nodes termed anchors and inter-node bearings measured in a common reference frame. The contributions of the paper are twofold. Firstly, we propose necessary and sufficient conditions for network localizability with both algebraic and rigidity theoretic interpretations. The analysis of the localizability heavily relies on the recently developed bearing rigidity theory and a special matrix termed the bearing Laplacian. Secondly, we propose a linear distributed protocol for bearing-based network localization. The protocol can globally localize a network if and only if the network is localizable. The sensitivity of the protocol to constant measurement errors is also analyzed. One novelty of this work is that the localizability analysis and localization protocol are applicable to networks in arbitrary dimensional spaces.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Bearing Rigidity and Almost Global Bearing-Only Formation Stabilization

A fundamental problem that the bearing rigidity theory studies is to determine when a framework can be uniquely determined up to a translation and a scaling factor by its inter-neighbor bearings. While many previous works focused on the bearing rigidity of two-dimensional frameworks, a first contribution of this paper is to extend these results to arbitrary dimensions. It is shown that a framework in an arbitrary dimension can be uniquely determined up to a translation and a scaling factor by the bearings if and only if the framework is infinitesimally bearing rigid. In this paper, the proposed bearing rigidity theory is further applied to the bearing-only formation stabilization problem where the target formation is defined by inter-neighbor bearings and the feedback control uses only bearing measurements. Nonlinear distributed bearing-only formation control laws are proposed for the cases with and without a global orientation. It is proved that the control laws can almost globally stabilize infinitesimally bearing rigid formations. Numerical simulations are provided to support the analysis.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Bearing-Based Distributed Control and Estimation of Multi-Agent Systems

This paper studies the distributed control and estimation of multi-agent systems based on bearing information. In particular, we consider two problems: (i) the distributed control of bearing-constrained formations using relative position measurements and (ii) the distributed localization of sensor networks using bearing measurements. Both of the two problems are considered in arbitrary dimensional spaces. The analyses of the two problems rely on the recently developed bearing rigidity theory. We show that the two problems have the same mathematical formulation and can be solved by identical protocols. The proposed controller and estimator can globally solve the two problems without ambiguity. The results are supported with illustrative simulations.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Bearing-Based Formation Maneuvering

This paper studies the problem of multi-agent formation maneuver control where both of the centroid and scale of a formation are required to track given velocity references while maintaining the formation shape. Unlike the conventional approaches where the target formation is defined by inter-neighbor relative positions or distances, we propose a bearing-based approach where the target formation is defined by inter-neighbor bearings. Due to the invariance of the bearings, the bearing-based approach provides a natural solution to formation scale control. We assume the dynamics of each agent as a single integrator and propose a globally stable proportional-integral formation maneuver control law. It is shown that at least two leaders are required to collaborate in order to control the centroid and scale of the formation whereas the followers are not required to have access to any global information, such as the velocities of the leaders.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Bearing-Based Formation Stabilization with Directed Interaction Topologies

This paper studies the problem of stabilizing target formations specified by inter-neighbor bearings with relative position measurements. While the undirected case has been studied in the existing works, this paper focuses on the case where the interaction topology is directed. It is shown that a linear distributed control law, which was proposed previously for undirected cases, can still be applied to the directed case. The formation stability in the directed case, however, relies on a new notion termed bearing persistence, which describes whether or not the directed underlying graph is persistent with the bearing rigidity of a formation. If a target formation is not bearing persistent, undesired equilibriums will appear and global formation stability cannot be guaranteed. The notion of bearing persistence is defined by the bearing Laplacian matrix and illustrated by simulation examples.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Translational and Scaling Formation Maneuver Control via a Bearing-Based Approach

This paper studies distributed maneuver control of multi-agent formations in arbitrary dimensions. The objective is to control the translation and scale of the formation while maintaining the desired formation pattern. Unlike conventional approaches where the target formation is defined by relative positions or distances, we propose a novel bearing-based approach where the target formation is defined by inter-neighbor bearings. Since the bearings are invariant to the translation and scale of the formation, the bearing-based approach provides a simple solution to the problem of translational and scaling formation maneuver control. Linear formation control laws for double-integrator dynamics are proposed and the global formation stability is analyzed. This paper also studies bearing-based formation control in the presence of practical problems including input disturbances, acceleration saturation, and collision avoidance. The theoretical results are illustrated with numerical simulations.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

Finite-time Stabilization of Circular Formations using Bearing-only Measurements

This paper studies decentralized formation control of multiple vehicles when each vehicle can only measure the local bearings of their neighbors by using bearing-only sensors. Since the inter-vehicle distance cannot be measured, the target formation involves no distance constraints. More specifically, the target formation considered in this paper is an angle-constrained circular formation, where each vehicle has exactly two neighbors and the angle at each vehicle subtended by its two neighbors is pre-specified. To stabilize the target formation, we propose a discontinuous control law that only requires the sign information of the angle errors. Due to the discontinuity of the proposed control law, the stability of the closed-loop system is analyzed by employing a locally Lipschitz Lyapunov function and nonsmooth analysis tools. We prove that the target formation is locally finite-time stable with collision avoidance guaranteed. The evolution of the vehicle positions in the plane is also characterized.

0 citations0 reviewsNo code linkedOpen access
preprint2012arXiv

Distributed Control of Angle-constrained Circular Formations using Bearing-only Measurements

This paper studies distributed formation control of multiple agents in the plane using bearing-only measurements. It is assumed that each agent only measures the local bearings of their neighbor agents. The target formation considered in this paper is a circular formation, where each agent has exactly two neighbors. In the target formation, the angle subtended at each agent by their two neighbors is specified. We propose a distributed control law that stabilizes angle-constrained target formations merely using local bearing measurements. The stability of the target formation is analyzed based on Lyapunov approaches. We present a unified proof to show that our control law not only can ensure local exponential stability but also can give local finite-time stability. The exponential or finite-time stability can be easily switched by tuning a parameter in the control law.

0 citations0 reviewsNo code linkedOpen access
preprint2012arXiv

Optimal Sensor Placement for Target Localization and Tracking in 2D and 3D

This paper analytically characterizes optimal sensor placements for target localization and tracking in 2D and 3D. Three types of sensors are considered: bearing-only, range-only, and received-signal-strength. The optimal placement problems of the three sensor types are formulated as an identical parameter optimization problem and consequently analyzed in a unified framework. Recently developed frame theory is applied to the optimality analysis. We prove necessary and sufficient conditions for optimal placements in 2D and 3D. A number of important analytical properties of optimal placements are further explored. In order to verify the analytical analysis, we present a gradient control law that can numerically construct generic optimal placements.

0 citations0 reviewsNo code linkedOpen access