arXiv subjects

Seungjae Lee

Publications and source records attributed to Seungjae Lee.

At least 19 recordsLinked to original sources

arXiv2026

HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos

Human egocentric video captures rich manipulation demonstrations without any robot hardware, yet transferring these skills to robots remains challenging due to the embodiment gap between human and robot in both visual appearance and kinematics. We present HumanEgo, a framework that bridges the embodiment gap by lifting each human demonstration to an entity-level representation of hand-object interaction, and training a flow matching policy with dense auxiliary objectives that amplify supervision from every trajectory. HumanEgo is robot-data-free, hardware-agnostic, data-efficient, and zero-shot human-to-robot transferable. With only 30 minutes of human videos per task, HumanEgo achieves 92.5% average success across four real-world tasks (75% with just 15 minutes), outperforms matched-time robot teleoperation by 41%, and robustly transfers zero-shot across novel robots, cameras, and environments. We release HumanEgo as an easy-to-use, open-source framework for learning robot policies directly from human data: https://github.com/TX-Leo/HumanEgo

cs.RO↗

arXiv2026

Beyond Episodic Evaluation: Memory Architectural Bottlenecks in Sequential Embodied Question Answering

Embodied question answering (EQA) is traditionally evaluated under an episodic formulation, where agents solve each task independently and reset internal state between episodes. However, real-world robots operate continuously and must accumulate, retain, and selectively reuse information acquired from prior interactions. Despite this practical requirement, the architectural mechanisms needed to support sequential memory in EQA remain underexplored. In this work, we investigate how different memory architectures behave when EQA agents are evaluated sequentially, with multiple questions answered in the same scene while memory is carried forward across queries. We find that simply preserving existing memory is often insufficient. Agents that retain only traversability information, such as 2D occupancy maps, remember where the robot has explored but not the visual-semantic evidence needed for later questions. Agents trained on short-horizon episodic data face a different challenge: when exposed to continuous, multi-query histories, their inherited context suffers from severe temporal mismatch, rather than forming a reusable scene representation. To overcome this architectural bottleneck, we highlight the necessity of structured, spatially grounded memory: architectures that map persistent visual observations onto metric 3D geometry preserve visual-semantic evidence in a coherent scene representation. Extensive experiments in simulated environments reveal that this form of memory breaks the accuracy-efficiency tradeoff in sequential settings, simultaneously achieving higher answer accuracy and lower navigation costs. We further validate these findings on a real-world mobile robot, demonstrating that spatially grounded visual memory is critical for enabling continuous, intelligent operation in physical environments.

cs.RO↗

arXiv2026

A combinatorial Approach to $α$-Ricci and Lin-Lu-Yau Ricci curvatures on Graphs

In this paper, we study the $α$-Ricci curvature and the Lin-Lu-Yau Ricci curvature on simple, connected, and locally finite graphs. For regular graphs, we introduce a combinatorial construction of optimal transport plans realizing the 1-Wasserstein distance and use it to derive exact formulas for the $α$-Ricci curvature and the Lin-Lu-Yau Ricci curvature. This yields a combinatorial proof of the known curvature formulas. Furthermore, for non-regular graphs, we characterize conditions on the size of common neighborhoods that guarantee either non-negative or vanishing Lin-Lu-Yau Ricci curvature.

math.DG↗

arXiv2026

Intermediate Pseudoconvexity of Fiber Bundles

In this paper, we investigate the pseudoconvexity of locally trivial holomorphic ball bundles over compact Riemann surfaces of genus $\geq 2$, as well as the intermediate pseudoconvexity of their complements in the associated projective space bundles. Inspired by Brunella's work, we prove that any such ball bundle is $1$-convex, while its complement is $n$-convex, where $n$ denotes the dimension of the ball fiber, provided that the bundle admits a harmonic section with a regular point.

math.CV↗

arXiv2026

DASIP: Dynamic Test-Time Compute Scaling for Robot Control with Stochastic Interpolant Policies

Diffusion- and flow-based policies deliver state-of-the-art performance on long-horizon robotic manipulation and imitation learning tasks. However, these controllers employ a fixed inference budget at every control step, regardless of task complexity, leading to computational inefficiency for simple subtasks while potentially underperforming on challenging ones. To address these issues, we introduce Difficulty-Aware Stochastic Interpolant Policy (DA-SIP), a framework that enables robotic controllers to adaptively adjust their integration horizon in real time based on task difficulty. Our approach employs a difficulty classifier that analyzes observations to dynamically select the step budget, the optimal solver variant, and ODE/SDE integration at each control cycle. DA-SIP builds upon the stochastic interpolant formulation to provide a unified framework that unlocks diverse training and inference configurations for diffusion- and flow-based policies. Through comprehensive benchmarks across diverse manipulation tasks, DA-SIP achieves 2.6-4.4x reduction in total computation time while maintaining task success rates comparable to fixed maximum-computation baselines. By implementing adaptive computation within this framework, DA-SIP transforms generative robot controllers into efficient, task-aware systems that intelligently allocate inference resources where they provide the greatest benefit.

cs.RO↗

arXiv2026

$μ_0$: A Scalable 3D Interaction-Trace World Model

World models that capture how actions induce physical change enable scalable robot learning without reliance on embodiment-specific action labels. Pixel-space video models provide broad visual priors but expend model capacity on dense appearance reconstruction, while direct action models require embodiment-specific labels that hinder scalability. We present $μ_0$, a scalable world model based on 3D traces. Rather than predicting dense pixels or directly modeling actions, $μ_0$ forecasts smooth 3D trajectories for salient interaction points such as objects, tools, hands, and contact regions, yielding a compact, embodiment-agnostic motion interface. To enable training from diverse video sources, our TraceExtract system automatically extracts 3D supervision by selecting keypoints, constructing globally aligned traces, and associating motion segments with hierarchical language captions. This TraceExtract supervision pretrains $μ_0$ by combining a pretrained vision-language backbone with a modular trace expert, which represents each query via B-spline control points and predicts future traces. Experiments show that $μ_0$ outperforms baselines in both 2D and 3D trace prediction, including trace prediction models and tokenized VLM methods. Because $μ_0$ is frozen and reusable, it can be paired with action experts for downstream robot embodiments. Despite action-free pretraining, the resulting trace-conditioned policies achieve performance competitive with VLA models pretrained with action supervision, such as $π_0$. These results establish 3D traces as a scalable and transferable representation for cross-embodiment manipulation.

cs.RO↗

arXiv2026

DynaFLIP: Rethinking Robotics Perception via Tri-Modal-Dynamics Guided Representation

Robot manipulation critically depends on perception that preserves the action-relevant aspects of a scene. Yet most robot learning pipelines are built upon visual encoders pre-trained for static recognition or vision-language alignment, leaving motion understanding to downstream policies. We introduce DynaFLIP, a dynamics-aware multimodal pre-training framework that pushes motion understanding upstream into perception. We construct image-language-3D flow triplets from heterogeneous human and robot videos, and use these triplets as training-time supervision to shape an image-only encoder. Our key idea is to encourage the three modalities to span a small simplex volume in the shared hyperspherical space -- a smaller simplex volume indicating stronger alignment. To avoid the geometric ambiguity and trivial collapse of naive volume minimization, we combine simplex-volume minimization with a cosine regularizer and a contrastive objective. Our analyses show that DynaFLIP focuses on control-relevant regions critical for manipulation. The resulting dynamics-aware representations serve as reusable visual backbones and consistently outperform baselines across diverse downstream policies, including VLAs. We validate this across diverse simulation and real-world setups, with gains reaching +22.5% under out-of-distribution scenarios. Our results suggest that robot generalization improves when visual representations are trained to encode not just what is present, but how the world changes under action.

cs.RO↗

arXiv2026

Raymoval: Raycasting-based Dynamic Object Removal for Static 3D Mapping

Static mapping is fundamental to robot navigation, providing a persistent geometric prior and a consistent reference for long-term autonomy. However, dynamic objects leave residual traces and cause surface loss, which reduces map consistency. We propose a raycasting-based module for dynamic object removal in static 3D mapping. Each scan is projected onto an azimuth-elevation grid, and for every viewing direction we compare the bin-wise minimum range with the map's first-hit distance computed by raycasting. Furthermore, we apply a raycast consistency test that separates dynamic from static points. Finally, a spatial consistency validation step refines labels, producing static maps with lower residual dynamics and reduced over-removal. We evaluate our approach quantitatively and qualitatively on SemanticKITTI and a challenging custom dataset, and show consistent static mapping results.

cs.RO↗

arXiv2026

SE(3)-LIO: Smooth IMU Propagation With Jointly Distributed Poses on SE(3) Manifold for Accurate and Robust LiDAR-Inertial Odometry

In estimating odometry accurately, an inertial measurement unit (IMU) is widely used owing to its high-rate measurements, which can be utilized to obtain motion information through IMU propagation. In this paper, we address the limitations of existing IMU propagation methods in terms of motion prediction and motion compensation. In motion prediction, the existing methods typically represent a 6-DoF pose by separating rotation and translation and propagate them on their respective manifold, so that the rotational variation is not effectively incorporated into translation propagation. During motion compensation, the relative transformation between predicted poses is used to compensate motion-induced distortion in other measurements, while inherent errors in the predicted poses introduce uncertainty in the relative transformation. To tackle these challenges, we represent and propagate the pose on SE(3) manifold, where propagated translation properly accounts for rotational variation. Furthermore, we precisely characterize the relative transformation uncertainty by considering the correlation between predicted poses, and incorporate this uncertainty into the measurement noise during motion compensation. To this end, we propose a LiDAR-inertial odometry (LIO), referred to as SE(3)-LIO, that integrates the proposed IMU propagation and uncertainty-aware motion compensation (UAMC). We validate the effectiveness of SE(3)-LIO on diverse datasets. Our source code and additional material are available at: https://se3-lio.github.io/.

cs.RO↗

arXiv2026

RESYSTANCE: Unleashing Hidden Performance of Compaction in LSM-trees via eBPF

The development of high-speed storage devices such as NVMe SSDs has shifted the primary I/O bottleneck from hardware to software. Modern database systems also rely on kernel-based I/O paths, where frequent system call invocations and kernel-user space transitions lead to relatively large overheads and performance degradation. This issue is particularly pronounced in Log-Structured Merge-tree (LSM-tree)-based NoSQL databases. We identified that, in particular, the background compaction process generates a large number of read system calls, causing significant overhead. To address this problem, we propose RESYSTANCE, which leverages eBPF and io_uring to free compaction from system calls and unlock hidden performance potential. RESYSTANCE improves disk I/O efficiency during read operations via io uring and significantly reduces software stack overhead by handling compaction directly inside the kernel through eBPF. Moreover, RESYSTANCE minimizes user-kernel transitions by offloading key I/O routines into the kernel without modifying the LSM-tree structure or compaction algorithm. RESYSTANCE was extensively evaluated using db_bench, YCSB, and OLTP workloads. Compared to baseline RocksDB, it reduced the average number of system call invocations during compaction by 99% and shortened compaction time by 50%. Consequently, in write-intensive workloads, RESYSTANCE improved throughput by up to 75% and reduced the p99 latency by 40%.

cs.DB↗

arXiv2026

MomaGraph: State-Aware Unified Scene Graphs with Vision-Language Model for Embodied Task Planning

Mobile manipulators in households must both navigate and manipulate. This requires a compact, semantically rich scene representation that captures where objects are, how they function, and which parts are actionable. Scene graphs are a natural choice, yet prior work often separates spatial and functional relations, treats scenes as static snapshots without object states or temporal updates, and overlooks information most relevant for accomplishing the current task. To address these limitations, we introduce MomaGraph, a unified scene representation for embodied agents that integrates spatial-functional relationships and part-level interactive elements. However, advancing such a representation requires both suitable data and rigorous evaluation, which have been largely missing. We thus contribute MomaGraph-Scenes, the first large-scale dataset of richly annotated, task-driven scene graphs in household environments, along with MomaGraph-Bench, a systematic evaluation suite spanning six reasoning capabilities from high-level planning to fine-grained scene understanding. Built upon this foundation, we further develop MomaGraph-R1, a 7B vision-language model trained with reinforcement learning on MomaGraph-Scenes. MomaGraph-R1 predicts task-oriented scene graphs and serves as a zero-shot task planner under a Graph-then-Plan framework. Extensive experiments demonstrate that our model achieves state-of-the-art results among open-source models, reaching 71.6% accuracy on the benchmark (+11.4% over the best baseline), while generalizing across public benchmarks and transferring effectively to real-robot experiments.

cs.CV↗

arXiv2025

TraceGen: World Modeling in 3D Trace Space Enables Learning from Cross-Embodiment Videos

Learning new robot tasks on new platforms and in new scenes from only a handful of demonstrations remains challenging. While videos of other embodiments - humans and different robots - are abundant, differences in embodiment, camera, and environment hinder their direct use. We address the small-data problem by introducing a unifying, symbolic representation - a compact 3D "trace-space" of scene-level trajectories - that enables learning from cross-embodiment, cross-environment, and cross-task videos. We present TraceGen, a world model that predicts future motion in trace-space rather than pixel space, abstracting away appearance while retaining the geometric structure needed for manipulation. To train TraceGen at scale, we develop TraceForge, a data pipeline that transforms heterogeneous human and robot videos into consistent 3D traces, yielding a corpus of 123K videos and 1.8M observation-trace-language triplets. Pretraining on this corpus produces a transferable 3D motion prior that adapts efficiently: with just five target robot videos, TraceGen attains 80% success across four tasks while offering 50-600x faster inference than state-of-the-art video-based world models. In the more challenging case where only five uncalibrated human demonstration videos captured on a handheld phone are available, it still reaches 67.5% success on a real robot, highlighting TraceGen's ability to adapt across embodiments without relying on object detectors or heavy pixel-space generation.

cs.RO↗

arXiv2025

Symmetric differentials and jets extension of $L^2$ holomorphic functions II: Explicit form

For a symmetric differential on the compact quotient $Σ= \mathbb{B}^n / Γ$ of the complex unit ball $\mathbb{B}^n \subset \mathbb{C}^n$ by a discrete subgroup $Γ\subset \mathrm{Aut}(\mathbb{B}^n)$, there exists a corresponding weighted $L^2$-holomorphic function on $(\mathbb{B}^n \times \mathbb{B}^n)/Γ$, where $Γ$ acts diagonally on $\mathbb{B}^n \times \mathbb{B}^n$. In this paper, we give an explicit description of this correspondence and derive several applications based on its explicit form.

math.CV↗

arXiv2025

LVI-Q: Robust LiDAR-Visual-Inertial-Kinematic Odometry for Quadruped Robots Using Tightly-Coupled and Efficient Alternating Optimization

Autonomous navigation for legged robots in complex and dynamic environments relies on robust simultaneous localization and mapping (SLAM) systems to accurately map surroundings and localize the robot, ensuring safe and efficient operation. While prior sensor fusion-based SLAM approaches have integrated various sensor modalities to improve their robustness, these algorithms are still susceptible to estimation drift in challenging environments due to their reliance on unsuitable fusion strategies. Therefore, we propose a robust LiDAR-visual-inertial-kinematic odometry system that integrates information from multiple sensors, such as a camera, LiDAR, inertial measurement unit (IMU), and joint encoders, for visual and LiDAR-based odometry estimation. Our system employs a fusion-based pose estimation approach that runs optimization-based visual-inertial-kinematic odometry (VIKO) and filter-based LiDAR-inertial-kinematic odometry (LIKO) based on measurement availability. In VIKO, we utilize the footpreintegration technique and robust LiDAR-visual depth consistency using superpixel clusters in a sliding window optimization. In LIKO, we incorporate foot kinematics and employ a point-toplane residual in an error-state iterative Kalman filter (ESIKF). Compared with other sensor fusion-based SLAM algorithms, our approach shows robust performance across public and longterm datasets.

cs.RO↗

arXiv2025

Sharp local well-posedness of $C^1$ vortex patches

It is well known that the boundary dynamics of vortex patches is globally well-posed in the Hölder space $C^{1,α}$ for $0<α<1$, whereas the well-posedness in $C^1$ remains an open problem, even locally. In this paper, we establish the local well-posedness for vortex patches in the space $C^{1,φ}$ defined via a modulus of continuity $φ$ that satisfies certain structural assumptions. Our class includes curves that are strictly rougher than the Hölder-continuous ones, with prototypical examples being $φ(r) = (-\log r)^{-s}$ for $s>3$. Motivated by the fact that the velocity operator in the contour dynamics equation is a nonlinear variant of the Hilbert transform, we study the system of equations satisfied by the curve parametrization $γ\in C^{1,φ}$ and its Hilbert transform. In doing so, we derive several properties of the Hilbert transform and its variants in critical spaces, which are essential for controlling the velocity operator and its Hilbert transform.

math.AP↗

arXiv2025

Why Are Web AI Agents More Vulnerable Than Standalone LLMs? A Security Analysis

Recent advancements in Web AI agents have demonstrated remarkable capabilities in addressing complex web navigation tasks. However, emerging research shows that these agents exhibit greater vulnerability compared to standalone Large Language Models (LLMs), despite both being built upon the same safety-aligned models. This discrepancy is particularly concerning given the greater flexibility of Web AI Agent compared to standalone LLMs, which may expose them to a wider range of adversarial user inputs. To build a scaffold that addresses these concerns, this study investigates the underlying factors that contribute to the increased vulnerability of Web AI agents. Notably, this disparity stems from the multifaceted differences between Web AI agents and standalone LLMs, as well as the complex signals - nuances that simple evaluation metrics, such as success rate, often fail to capture. To tackle these challenges, we propose a component-level analysis and a more granular, systematic evaluation framework. Through this fine-grained investigation, we identify three critical factors that amplify the vulnerability of Web AI agents; (1) embedding user goals into the system prompt, (2) multi-step action generation, and (3) observational capabilities. Our findings highlights the pressing need to enhance security and robustness in AI agent design and provide actionable insights for targeted defense strategies.

cs.LG↗

arXiv2025

Imagine, Verify, Execute: Memory-guided Agentic Exploration with Vision-Language Models

Exploration is essential for general-purpose robotic learning, especially in open-ended environments where dense rewards, explicit goals, or task-specific supervision are scarce. Vision-language models (VLMs), with their semantic reasoning over objects, spatial relations, and potential outcomes, present a compelling foundation for generating high-level exploratory behaviors. However, their outputs are often ungrounded, making it difficult to determine whether imagined transitions are physically feasible or informative. To bridge the gap between imagination and execution, we present IVE (Imagine, Verify, Execute), an agentic exploration framework inspired by human curiosity. Human exploration is often driven by the desire to discover novel scene configurations and to deepen understanding of the environment. Similarly, IVE leverages VLMs to abstract RGB-D observations into semantic scene graphs, imagine novel scenes, predict their physical plausibility, and generate executable skill sequences through action tools. We evaluate IVE in both simulated and real-world tabletop environments. The results show that IVE enables more diverse and meaningful exploration than RL baselines, as evidenced by a 4.1 to 7.8x increase in the entropy of visited states. Moreover, the collected experience supports downstream learning, producing policies that closely match or exceed the performance of those trained on human-collected demonstrations.

cs.RO↗

arXiv2025

Hopf-Induced Desynchronization

The emergence of synchrony essentially underlies the functionality of many systems across physics, biology and engineering. In all established synchronization phase transitions so far, a stable synchronous state is connected to a stable incoherent state: For continuous transitions, stable synchrony directly connects to stable incoherence at a critical point, whereas for discontinuous transitions, stable synchrony is connected to stable incoherence via an additional unstable branch. Here we present a novel type of transition between synchrony and incoherence where the synchronous state does not connect to the state of incoherence. We uncover such transitions in the complexified Kuramoto model with their variables and coupling strength parameter analytically continued. Deriving a self-consistency equation for a quaternion order parameter that we propose helps to mathematically pin down the mechanisms underlying this transition type. Local numerical analysis suggests that the transition is linked to a Hopf bifurcation destabilizing synchrony, in contrast to branching point bifurcations established for the transition between synchrony and incoherence so far.

nlin.AO↗