arXiv subjects

Hanbit Oh

Publications and source records attributed to Hanbit Oh.

At least 19 recordsLinked to original sources

arXiv2026

The Embodiment Gap in Robot Foundation Models

Robot foundation models (RFMs), including vision-language-action (VLA) policies, are often discussed through a scaling view: more data, larger models, and broader benchmarks should improve generalization. In robotics, however, a model can generalize while work still remains before it can run on a robot with a particular body. The work required differs across methods and target robots, and those differences affect practical deployment. We call the gap between reusable models, representations, or data and their use in execution on the target robot the embodiment gap. This survey examines what can be reused across robot embodiments and what must still be implemented on a new robot. We place existing methods on a two-axis map that shows the type of shared structure and the stage at which adaptation is needed for execution on the target robot. We then examine recent work through three overlapping research directions: sharing semantics and perception, sharing robot data and interfaces, and learning correspondence across embodiments. We also propose a reporting framework for adaptation work that success rate alone does not reveal. The framework identifies the work that should be checked when comparing cross-embodiment learning and highlights work that remains on a new robot and questions for future study.

cs.RO↗

arXiv2026

GuidedAttention: Interpretable and Correctable Visual Attention for OOD-Robust Robot Manipulation via Imitation Learning

End-to-end visuomotor policies provide little opportunity for humans to understand or correct the policy's visual attention. We propose GuidedAttention, a visuomotor imitation learning framework that introduces interpretable and correctable visual attention as an explicit intermediate representation. Task-relevant attention keypoints are predicted from camera images and condition a diffusion-based action policy. Users can inspect and optionally correct selected keypoints once at rollout initialization, after which the corrected attention is automatically propagated throughout execution by a tracking module. Experiments in simulation and the real world demonstrate that GuidedAttention consistently improves robot manipulation performance, particularly under positional and appearance out-of-distribution (OOD) conditions. https://mmurooka.github.io/guided-attention-project-page

cs.RO↗

arXiv2026

Incommensurate pair-density-wave correlations in two-leg ladder $t$--$J$--$J_\perp$ model

We report the discovery of a generalized Luther-Emery liquid phase characterized by incommensurate pair-density-wave (iC-PDW) correlations in the two-leg $t$-$J$-$J_\perp$ ladder model. By tuning the potential difference between the legs, we explore the regime of intermediate layer polarization $P$. Combining density-matrix renormalization group (DMRG) simulations with bosonization analysis, we identify a spin-gapped phase at finite $P$, where the interlayer and intralayer pair correlations both oscillate, but with distinct periodicities. The interlayer correlations exhibit FFLO-like oscillations, driven by pairing between layers with mismatched Fermi momenta, with a period determined by their momentum difference. In contrast, the intralayer pair correlations arise from the coupling between charges on one layer and spin fluctuations on the opposite layer, with a momentum equal to twice the Fermi momentum of the opposite layer. The iC-PDW state is robust across a wide range of doping and polarization, although finite interlayer hopping eventually destabilizes it toward a state with charge-$4e$ correlations. We conclude by discussing the experimental realization of this model in optical lattice platforms and its relevance to the bilayer nickelate La$_3$Ni$_2$O$_7$.

cond-mat.str-el↗

arXiv2025

Pair-density-wave superconductivity and Anderson's theorem in bilayer nickelates

The recent experimental observations of high temperature superconductivity in bilayer nickelate have attracted lots of attentions. Previous studies have assumed a mirror symmetry $\mathcal M$ between the two layers and focused on uniform and clean superconducting states. Here, we show that breaking this mirror symmetry via an applied displacement field can stabilize a pair-density-wave (PDW) superconductor, which is similar to the Fulde--Ferrell--Larkin--Ovchinnikov (FFLO) state, but at zero magnetic field. Based on a mean-field analysis of a model of $d_{x^2-y^2}$ orbital with an effective inter-layer attraction, we demonstrate that the PDW phase is robust over a wide range of displacement field, interlayer hopping strengths, and electron fillings. Finally, we analyze disorder effects on interlayer superconductivity within the first Born approximation. Based on symmetry considerations, we show that pairing is weaken by disorders which break the mirror symmetry, even with unbroken time reversal symmetry. Our results establish bilayer nickelate as a tunable platform for realizing finite-momentum pairing and for exploring generalized disorder effects.

cond-mat.str-el↗

arXiv2025

Origin of Spin Stripes in Bilayer Nickelate La$_3$Ni$_2$O$_7$

The bilayer nickelate La$_3$Ni$_2$O$_7$ has recently emerged as a high-temperature superconductor with unusual spin stripe order in its ambient pressure phase. We propose a microscopic Hamiltonian that faithfully reflects the crystalline symmetry of this system, with the primary aim of addressing its unconventional magnetism at ambient pressure. Using state-of-the-art density matrix renormalization group calculations, we show that $(\pi/2,\pi/2)$ spin stripe order arises in our model at sizable Hund's coupling $J_H$ from a hidden quasi-one-dimensionality and persists over a range of electron concentrations. In the more symmetric high-pressure regime, our model exhibits enhanced interlayer pairing tendencies when the interlayer antiferromagnetic coupling $J_{\bot}$ becomes sufficiently large. Our results provide a microscopic origin of the diagonal spin stripes and identify Hund's coupling $J_H$ and interlayer coupling $J_{\bot}$ as key ingredients governing magnetic order and pairing tendencies in La$_3$Ni$_2$O$_7$.

cond-mat.supr-con↗

arXiv2025

RoboManipBaselines: A Unified Framework for Imitation Learning in Robotic Manipulation across Real and Simulation Environments

We present RoboManipBaselines, an open-source software framework for imitation learning research in robotic manipulation. The framework supports the entire imitation learning pipeline, including data collection, policy training, and rollout, across both simulation and real-world environments. Its design emphasizes integration through a consistent workflow, generality across diverse environments and robot platforms, extensibility for easily adding new robots, tasks, and policies, and reproducibility through evaluations using publicly available datasets. RoboManipBaselines systematically implements the core components of imitation learning: environment, dataset, and policy. Through a unified interface, the framework supports multiple simulators and real robot environments, as well as multimodal sensors and a wide variety of policy models. We further present benchmark evaluations in both simulation and real-world environments and introduce several research applications, including data augmentation, integration with tactile models, interactive robotic systems, 3D sensing evaluation, and hardware extensions. These results demonstrate that RoboManipBaselines provides a useful foundation for advancing research and experimental validation in robotic manipulation using imitation learning. https://isri-aist.github.io/RoboManipBaselines-ProjectPage

cs.RO↗

arXiv2025

Self-Augmented Robot Trajectory: Efficient Imitation Learning via Safe Self-augmentation with Demonstrator-annotated Precision

Imitation learning is a promising paradigm for training robot agents; however, standard approaches typically require substantial data acquisition -- via numerous demonstrations or random exploration -- to ensure reliable performance. Although exploration reduces human effort, it lacks safety guarantees and often results in frequent collisions -- particularly in clearance-limited tasks (e.g., peg-in-hole) -- thereby, necessitating manual environmental resets and imposing additional human burden. This study proposes Self-Augmented Robot Trajectory (SART), a framework that enables policy learning from a single human demonstration, while safely expanding the dataset through autonomous augmentation. SART consists of two stages: (1) human teaching only once, where a single demonstration is provided and precision boundaries -- represented as spheres around key waypoints -- are annotated, followed by one environment reset; (2) robot self-augmentation, where the robot generates diverse, collision-free trajectories within these boundaries and reconnects to the original demonstration. This design improves the data collection efficiency by minimizing human effort while ensuring safety. Extensive evaluations in simulation and real-world manipulation tasks show that SART achieves substantially higher success rates than policies trained solely on human-collected demonstrations. Video results available at https://sites.google.com/view/sart-il .

cs.RO↗

arXiv2025

Doping a spin-one Mott insulator: possible application to bilayer nickelate

In this article, we review some recent theoretical developments on potential high-temperature superconductors and unconventional metallic states that can arise from doping a spin-one Mott insulator in the $d^{8}$ valence. These studies are particularly relevant-though not limited-to the recently discovered bilayer nickelate superconductor La$_3$Ni$_2$O$_7$. We focus on a ferromagnetic (FM) Kondo lattice model with mobile electrons in the $d_{x^2-y^2}$ orbital coupled to the localized spin moments in $d_{z^2}$ orbital through a large Hund's coupling $J_H$. In the large $J_H$ limit, the model reduces to the type II t-J model with a mixture of spin-half singlon states and spin-one doublon states. We summarize DMRG results on the Luther-Emery liquid in one dimensional chain and two-leg ladder. Then we mainly focus on bilayer square lattice and show that a large inter-layer coupling $J_\perp$ of $d_{z^2}$ orbital can induce strong inter-layer pairing of $d_{x^2-y^2}$ orbital. In the strong $J_\perp$ limit, a kinetic-energy driven high $T_c$ superconductivity is demonstrated in an ideal model with only a single hopping term. Furthermore, the model predicts a symmetric pseudogap metal-dubbed `second Fermi liquid"-in the underdoped regime, yielding a phase diagram analogous to that of hole-doped cuprates. The bilayer Kondo model therefore, presents a promising platform for both realizing higher-Tc superconductors and exploring non-Fermi liquid physics. We also comment on the possible limitations of the current models for the bilayer nickelate material and point out some future directions.

cond-mat.str-el↗

arXiv2025

High spin, low spin or gapped spins: magnetism in the bilayer nickelates

Inspired by the recent discovery of high-temperature superconductivity in bilayer nickelates, we investigate the role of magnetism emerging from a hypothetical insulating $d^8$ parent state. We demonstrate that due to the interplay of superexchange and Hund's coupling, the system can be in a high-spin, low-spin or spin-gapped state. The low-spin state has singlets across the bilayer in the $d_{z^2}$ orbital, with charge carriers in the $d_{x^2-y^2}$ orbital. Thus, at low energy scales, it behaves as an effective one band system when hole doped. By contrast, the high-spin state is a more robust, spin-1 antiferromagnet. Using Hartree-Fock methods, we find that for fixed interaction strength and doping, high-spin magnetism remains more robust than the low-spin counterpart. Whether this implies that the high spin state provides a stronger pairing glue, or more strongly competes with superconductivity remains an open question. Our analysis therefore underscores the importance of identifying the spin state for understanding superconductivity in nickelates.

cond-mat.str-el↗

arXiv2025

Robust Instant Policy: Leveraging Student's t-Regression Model for Robust In-context Imitation Learning of Robot Manipulation

Imitation learning (IL) aims to enable robots to perform tasks autonomously by observing a few human demonstrations. Recently, a variant of IL, called In-Context IL, utilized off-the-shelf large language models (LLMs) as instant policies that understand the context from a few given demonstrations to perform a new task, rather than explicitly updating network models with large-scale demonstrations. However, its reliability in the robotics domain is undermined by hallucination issues such as LLM-based instant policy, which occasionally generates poor trajectories that deviate from the given demonstrations. To alleviate this problem, we propose a new robust in-context imitation learning algorithm called the robust instant policy (RIP), which utilizes a Student's t-regression model to be robust against the hallucinated trajectories of instant policies to allow reliable trajectory generation. Specifically, RIP generates several candidate robot trajectories to complete a given task from an LLM and aggregates them using the Student's t-distribution, which is beneficial for ignoring outliers (i.e., hallucinations); thereby, a robust trajectory against hallucinations is generated. Our experiments, conducted in both simulated and real-world environments, show that RIP significantly outperforms state-of-the-art IL methods, with at least $26\%$ improvement in task success rates, particularly in low-data scenarios for everyday tasks. Video results available at https://sites.google.com/view/robustinstantpolicy.

cs.RO↗

arXiv2024

High-temperature superconductivity from kinetic energy

Superconductivity is usually assumed to arise from attractive interaction. In this work we show that strong pairing is possible soley from kinetic energy even without a net attraction. We demonstrate a high-temperature kinetic superconductor in a simple lattice model with nearest-neighbor hopping ($t$) projected onto a constrained Hilbert space, analogous to the $t$-$J$ model with $J=0$, where kinetic magnetism has been previously studied. Using density matrix renormalization group (DMRG) on cylinders up to width $L_y=8$, we find a superconducting ground state exhibiting a key difference from high-$T_c$ cuprates: both the pairing gap and phase stiffness \textit{increase} with doping ($x$). We find pairing gaps, determined from spin and single-electron charge gaps, exceeding $1.5t$. This model can be realized within the double Kondo lattice model, relevant to bilayer nickelates, in the limit of strong inter-layer spin coupling ($J_\perp/t \rightarrow +\infty$) and a balancing inter-layer repulsion ($V$). Importantly, the double Kondo model does not fundamentally restrict $J_\perp/t$, suggesting the potential for high critical temperatures ($T_c$) approaching $0.5t$. While this idealized limit predicts large pairing gaps, we show a smooth connection to the more realistic regime with $J_\perp \sim t$, albeit with a reduced pairing gap of approximately $0.1t$. Assuming $t \sim 10^3$ K in typical solid state systems, our model suggests the exciting possibility of achieving $T_c$ of hundreds of Kelvin. We propose searching for bilayer materials with reduced out-of-plane lattice constants to better approximate the conditions of our ideal model.

cond-mat.str-el↗

arXiv2024

Strong pairing and symmetric pseudogap metal in double Kondo lattice model: from nickelate superconductor to tetralayer optical lattice

In this work, we propose and study a double Kondo lattice model which hosts robust superconductivity. The system consists of two identical Kondo lattice model, each with Kondo coupling $J_K$ within each layer, while the localized spin moments are coupled together via an inter-layer on-site antiferromagnetic spin coupling $J_\perp$. We consider the strong $J_\perp$ limit, wherein the local moments tend to form rung singlets and are thus gapped. However, the Kondo coupling $J_K$ transmits the inter-layer entanglement between the local moments to the itinerant electrons. Consequently, the itinerant electrons experience a strong inter-layer antiferromangetic spin coupling and form strong inter-layer pairing, which is confirmed through numerical simulation in one dimensional system. Experimentally, the $J_K \rightarrow -\infty$ limits of the model describes the recently found bilayer nickelate La$_3$Ni$_2$O$_7$, while the $J_K>0$ side can be realized in tetralayer optical lattice of cold atoms. Two extreme limits, $J_K \rightarrow -\infty$ and $J_K \rightarrow +\infty$ limit are shown to be simplified to a bilayer type II t-J model and a bilayer one-orbital t-J model, respectively. Thus, our double Kondo lattice model offers a unified framework for nickelate superconductor and tetralayer optical lattice quantum simulator upon changing the sign of $J_K$. We highlight both the qualitative similarity and the quantitative difference in the two sides of $J_K$. Finally, we discuss the possibility of a symmetric Kondo breakdown transition in the model with a symmetric pseudogap metal corresponding to the usual heavy Fermi liquid.

cond-mat.str-el↗

arXiv2024

Type II t-J model in charge transfer regime in bilayer La$_3$Ni$_2$O$_7$ and trilayer La$_4$Ni$_3$O$_{10}$

Recent observations of an 80 K superconductor in La$_3$Ni$_2$O$_7$ under high pressure have attracted significant attention. Recent experiments indicate that La$_3$Ni$_2$O$_7$ may be in the charge transfer regime, challenging the previous models based purely on the Ni $d_{x^2-y^2}$ and $d_{z^2}$ orbitals. In this study, we propose a low energy model that incorporates doped holes in the oxygen $p$ orbitals. Given that the parent nickel state is in the $3d^{8}$ configuration with a spin-one moment, doped hole only screens it down to spin-half, in contrast to the Zhang-Rice singlet in cuprate. We dub the single hole state as Zhang-Rice spin-half and build an effective model which includes three spin-one states ($d^8$) and two Zhang-Rice spin-half states ($d^8 L$). At moderate pressure around $20$ GPa, the dominated oxygen orbital is an in-plane Wannier orbital with the same lattice symmetry as the $d_{x^2-y^2}$ orbital. The resulting model reduces to the bilayer type II t-J model previously proposed in the Mott-Hubbard regime. Notably, the hopping between the in-plane $p$ orbitals of the two layers is still suppressed. Density matrix renormalization group (DMRG) simulation reveals a pairing dome with the optimal hole doping level at $x=0.4\sim0.5$, distinct from the hole doped cuprate where optimal doping occurs around $x=0.19$. Further increasing pressure initially raises the critical temperature ($T_c$) until reaching an optimal pressure beyond which the $p_z$ orbital of oxygen becomes favorable and superconductivity is diminished. This shift from in-plane $p$ orbital to $p_z$ orbital may elucidate the experimentally observed superconducting dome with varying pressure. As an extension, we also suggest a trilayer version of the type II t-J model as the minimal model for pressured La$_4$Ni$_3$O$_{10}$, which is distinct from the models in the Mott-Hubbard regime.

cond-mat.str-el↗

arXiv2024

Leveraging Demonstrator-perceived Precision for Safe Interactive Imitation Learning of Clearance-limited Tasks

Interactive imitation learning is an efficient, model-free method through which a robot can learn a task by repetitively iterating an execution of a learning policy and a data collection by querying human demonstrations. However, deploying unmatured policies for clearance-limited tasks, like industrial insertion, poses significant collision risks. For such tasks, a robot should detect the collision risks and request intervention by ceding control to a human when collisions are imminent. The former requires an accurate model of the environment, a need that significantly limits the scope of IIL applications. In contrast, humans implicitly demonstrate environmental precision by adjusting their behavior to avoid collisions when performing tasks. Inspired by human behavior, this paper presents a novel interactive learning method that uses demonstrator-perceived precision as a criterion for human intervention called Demonstrator-perceived Precision-aware Interactive Imitation Learning (DPIIL). DPIIL captures precision by observing the speed-accuracy trade-off exhibited in human demonstrations and cedes control to a human to avoid collisions in states where high precision is estimated. DPIIL improves the safety of interactive policy learning and ensures efficiency without explicitly providing precise information of the environment. We assessed DPIIL's effectiveness through simulations and real-robot experiments that trained a UR5e 6-DOF robotic arm to perform assembly tasks. Our results significantly improved training safety, and our best performance compared favorably with other learning methods.

cs.RO↗

arXiv2023

Strong pairing from small Fermi surface beyond weak coupling: Application to La$_3$Ni$_2$O$_7$

The studies of high-temperature superconductors raise a fundamental question: Can a small Fermi surface phase, which violates the Luttinger theorem, exist and give rise to superconductivity? Our work provides a positive answer through a controlled theory based on a bilayer model with strong inter-layer spin-spin coupling ($J_\perp$) but no inter-layer hopping ($t_\perp$). Then small hole doping of the rung-singlet insulator with two electrons per rung naturally leads to small hole pockets with Fermi surface volume per flavor smaller than the free fermion result by $1/2$ of the Brillouin zone(BZ). We construct a new t-J model on a bilayer square lattice, so called ESD t-J model and employ a generalized slave boson theory, which captures this small Fermi surface phase at small hole doping $x$. This metallic state is an intrinsically strongly correlated Fermi liquid beyond weak coupling theory, violating the perturbative Luttinger theorem but consistent with the Oshikawa's non-perturbative proof. We further show that it transitions into an inter-layer paired $s'$-wave superconductor at lower temperature through Feshbach resonance with a virtual Cooper pair, with a surprising doping-induced crossover from Bardeen-Cooper-Schrieffer (BCS) to Bose-Einstein condensation (BEC) at higher hole doping levels. This leads to a superconducting dome centered around $x=0.5$, with the normal state changing from the conventional Fermi liquid in the $x>0.5$ to the unusual small Fermi surface state in the $x<0.5$ side. Our theoretical findings including phase diagrams are also confirmed by density matrix renormalization group (DMRG) simulation in quasi one dimension. Applying our theoretical framework, we provide a plausible scenario for the recently found nickelate La$_3$Ni$_2$O$_7$ materials.

cond-mat.str-el↗

arXiv2023

S-wave pairing in a two-orbital t-J model on triangular lattice: possible application to Pb$_{10-x}$Cu$_x$(PO$_4$)$_6$O

Recently room temperature superconductor was claimed in Pb$_{10-x}$Cu$_x$(PO$_4$)$_6$O (also known as LK-99) with $x\in (0.9,1.1)$. Density functional theory (DFT) calculations suggest that the conduction electrons are from the doped Cu atoms with valence close to $d^{9}$. Motivated by this picture, we build a two-orbital Hubbard model on a triangular lattice formed by the $d_{xz}$ and $d_{yz}$ orbitals with total hole density (summed over spin and orbital) $n=1-p$. When $p=0$, the system is in a Mott insulator within this model. When $p>0$, we derive a $t-J$ model and perform a self-consistent slave boson mean field calculation. Interestingly we find a s-wave pairing in contrast to the one-orbital t-J model which favors $d+id$ pairing. S wave pairing should be more robust to disorder and may lead to high Tc superconductor with sufficiently large values of $t$ and $J$. However, the DFT calculations predict a very small value of $t$ and then the $T_c$ is expected to be small. If LK99 is really a high Tc superconductor, ingredients beyond the current model are needed. We conjecture that the doped Cu atoms may distort the original lattice and form local clusters with smaller Cu -Cu distance and thus larger values of $t$ and $J$. Within these clusters, we may locally apply our t-J model calculation and expect high Tc s-wave superconductor. Then the superconducting islands couple together, which may eventually become a global superconductor, an insulator or even an anomalous metal depending on sample details.

cond-mat.supr-con↗

arXiv2023

Type II t-J model and shared antiferromagnetic spin coupling from Hund's rule in superconducting La$_3$Ni$_2$O$_7$

Recently, a 80 K superconductor was discovered in La$_3$Ni$_2$O$_7$ under high pressure. Density function theory (DFT) calculations identify $d_{x^2-y^2}$, $d_{z^2}$ as the active orbitals on the bilayer square lattice with a $d^{8-x}$ configuration of of Ni per site. One naive expectation is to describe this system in terms of a two-orbital t-J model. However, we emphasize the importance of Hund's coupling $J_H$ and the $x=0$ limit should be viewed as a spin-one Mott insulator. Especially, the significant Hund's coupling shares the inter-layer super-exchange $J_\perp$ of the $d_{z^2}$ orbital to the $d_{x^2-y^2}$ orbital, an effect that cannot be captured by conventional perturbation or mean-field approaches. In this study, we first explore the limit where the $d_{z^2}$ orbital is Mott localized, dealing with a one-orbital bilayer t-J model focused on the $d_{x^2-y^2}$ orbital. Notably, we find that strong inter-layer pairing survives up to $x=0.5$ hole doping driven by the transmitted $J_\perp$, which explains the existence of a high Tc superconductor in the experiment at this doping level. Next, we uncover the more realistic situation where the $d_{z^2}$ orbital is slightly hole-doped and cannot be simply integrated out. We take the $J_H\rightarrow +\infty$ limit and propose a type II t-J model with four \textit{spin-half} singlon ($d^7$) states and three \textit{spin-one} doublon ($d^8$) states. Employing a parton mean-field approach, we recover similar results as in the one-orbital t-J model, but now with the effect of the $J_\perp$ automatically generated. We propose future experiments to electron dope the system to further enhance $T_c$.

cond-mat.str-el↗

arXiv2023

Disturbance Injection under Partial Automation: Robust Imitation Learning for Long-horizon Tasks

Partial Automation (PA) with intelligent support systems has been introduced in industrial machinery and advanced automobiles to reduce the burden of long hours of human operation. Under PA, operators perform manual operations (providing actions) and operations that switch to automatic/manual mode (mode-switching). Since PA reduces the total duration of manual operation, these two action and mode-switching operations can be replicated by imitation learning with high sample efficiency. To this end, this paper proposes Disturbance Injection under Partial Automation (DIPA) as a novel imitation learning framework. In DIPA, mode and actions (in the manual mode) are assumed to be observables in each state and are used to learn both action and mode-switching policies. The above learning is robustified by injecting disturbances into the operator's actions to optimize the disturbance's level for minimizing the covariate shift under PA. We experimentally validated the effectiveness of our method for long-horizon tasks in two simulations and a real robot environment and confirmed that our method outperformed the previous methods and reduced the demonstration burden.

cs.RO↗