arXiv subjects

Donghoon Kim

Publications and source records attributed to Donghoon Kim.

At least 19 recordsLinked to original sources

arXiv2026

HeLM: Hierarchical Explicit Language Memory for Non-Markovian Manipulation

Recent advances in Vision-Language-Action (VLA) models have enabled fine manipulation capabilities, yet performing tasks that require working memory remains a key challenge. Most existing VLA architectures operate under a Markovian assumption, mapping only the current observation to an action, which leads to failure in non-Markovian environments. Inspired by the human 'inner monologue', we propose Hierarchical Explicit Language Memory (HeLM), a hierarchical VLA framework that maintains task-relevant history in explicit language through factorized event detection and event-triggered memory updates. We evaluate HeLM across 10 non-Markovian real-world tasks with 56 variants, including five newly formulated inter-episode tasks that require persistent context across task boundaries. HeLM improves intra-episode success by 12.8 percentage points over the strongest baseline and achieves 78% subtask success on compositional generalization tasks, while using up to 4.5x less memory and running 2.5x faster than the keyframe baseline at long horizons.

cs.RO↗

arXiv2026

Universal purification dynamics in real non-unitary quantum processes

We study purification dynamics in monitored quantum processes governed by ensembles of quantum circuits in different random-matrix symmetry classes. We analyze the universal aspects that emerge away from the measurement induced phase transition and inside the volume/weak measurement phase and in the scaling limit of large time and Hilbert space dimension. We present two toy models that reveal two complementary visions and provide quantitative access to universal scaling: i) a discrete-time dynamic in which each time step corresponds to multiplication by a Gaussian random matrix; ii) weak continuous-time monitoring that induces a Dyson brownian motion of the eigenvalues of the density matrix. The first approach provides an algebraic characterization based on rotational invariance emerging in Kraus's operator space, focusing in particular on the unitary and orthogonal cases, respectively $β=2$ and $β=1$, with $β$ the Dyson random-matrix index. The second approach, on the other hand, allows for a unified treatment for any $β$, thanks to the mapping of the Fokker-Planck evolution of eigenvalues onto the Calogero-Sutherland integrable Hamiltonian diagonalized in terms of Jack polynomials. We provide explicit expressions for the universal decrease of Rényi entropies. We show that, approaching the universal scaling limit, numerical simulations of different models agree with each other and with our theoretical predictions. Our results clarify the existence of different classes of universality for the purification process in hybrid quantum systems, accessible in random circuit architectures and weak measurement protocols.

quant-ph↗

arXiv2026

Elicitation or Expansion? Demystifying Gains from Agent Harness Optimization

Recent studies in harness optimization have shown substantial performance gains in language model agents without updating model weights. A well-designed harness more effectively leverages the underlying model's capabilities by shaping context, tool use, and execution workflows. However, naïve success rates alone do not fully characterize whether optimized harness yields more reliable success on tasks already within reach (elicitation) or a broader range of solvable tasks (expansion). This distinction is crucial for recursive self-improvement and evolutionary search, as expansion may provide a broader range of successful experiences to push the capability boundaries. Despite the growing significance, this distinction remains underexplored in harness optimization. We conduct a systematic analysis to study the effect of gains from harness optimization. Building on pass@k analyses in the reinforcement learning (RL) literature that compare models before and after RL training, we compare optimized harnesses with their pre-optimization counterparts. Specifically, we assess changes in reachable task coverage, compare harness optimization with model scaling and increased inference-time compute, and examine how gains vary with task difficulty. Across static prediction and dynamic tasks, we find that pass@1 advantages of harness over stronger interventions can reverse at larger attempt budgets and aggregate gains conceal shortfalls on hard tasks, underscoring the need for a fine-grained diagnosis.

cs.LG↗

arXiv2026

MetaCurate: Evolving Curation Filters for High-Quality Agentic Coding Data

Supervised fine-tuning on agent trajectories is an effective way to improve coding agents, but selecting useful training examples requires more than identifying successful runs. A resolved patch may follow inefficient reasoning or unnecessary tool use, while a failed run may still contain useful intermediate steps. Training pipelines therefore rely on rules and qualitative assessments to filter trajectories, yet designing these filters typically requires substantial human expertise, and validating improvements through repeated training and evaluation is expensive. We introduce MetaCurate, a training-free framework that evolves executable trajectory curation filters using pairwise LLM judgments. Disagreements between the filter's rankings and the judge's preferences provide targeted feedback for jointly revising its rules, rubric, and aggregation policy. The resulting filter scores trajectories independently, so pairwise comparisons are required only during search. Starting from a blank filter, MetaCurate autonomously discovers a curation strategy that achieves competitive downstream performance without task-specific handwritten criteria. We validate downstream utility through controlled SFT experiments on our uncurated trajectory pool and the Nemotron v3 SWE dataset, with random selection and hand-authored filtering as baselines. In our experiments, MetaCurate achieves the highest mean SWE-Verified score among the compared methods, including hand-authored rule-based filtering and a training-loop baseline whose search costs 3.5x as much.

cs.LG↗

arXiv2026

Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following

Verifiable instruction-following benchmarks emphasize output-constraint satisfaction but can overlook whether responses meaningfully answer the underlying request, creating an evaluation blind spot. We show that frontier models can receive full credit for responses with little substantive answer content. Using the same verifier-based criterion as a training reward can induce reward hacking, as publicly released checkpoints exhibit higher constraint satisfaction but lower response quality after reinforcement learning with verifiable rewards. To address this reward hacking, we introduce CALIF, a lightweight method that augments verifier rewards with a calibrated likelihood-based bonus to preserve response quality. CALIF evaluates response likelihood under the request alone using the existing frozen reference policy, requiring no additional judge or learned reward model. Across models ranging from 1.7B to 8B in reasoning and non-reasoning modes, CALIF improves constraint satisfaction on IFBench while maintaining response quality near or above pre-RL levels. It also outperforms corresponding base policies in 40 of 45 model-metric comparisons across three rubric-based instruction-following benchmarks, demonstrating gains beyond training verifiers.

cs.LG↗

arXiv2026

Collectivity limits quantum entanglement

Understanding what limits many-body quantum entanglement is a central problem in physics. Spatial locality has long provided a fundamental mechanism: correlations between a region and its complement must be mediated through a small spatial interface, thereby constraining their entanglement. Here we show that universal constraints on entanglement can persist even when interactions are strongly nonlocal, through collectivity: many weak interactions generate fluctuations controlled by their square-summed, rather than total, strength. We establish this mechanism rigorously for generic gapped Hamiltonians with Kac-normalized power-law interactions $r^{-α}$ on a $D$-dimensional lattice. For arbitrary bipartitions, we prove that the ground-state entanglement scales at most logarithmically with system size for $α<D/2$ and subextensively for $D/2<α<D$, due to suppressed collective fluctuations around individual sites. For spatially regular bipartitions with codimension-one boundaries, we use a renormalization-group construction to extend this suppression to larger length scales, yielding polylogarithmic scaling for $D/2 < α< (D+1) / 2$ and parametrically stronger subvolume bounds for $(D+1)/2<α<D$. We also establish the corresponding optimality results: for arbitrary bipartitions, the logarithmic scaling for $α<D/2$ is optimal and $α=D/2$ marks the optimal threshold for universal logarithmic bounds, while for regular bipartitions the threshold at $α=(D+1)/2$ is likewise optimal for universal polylogarithmic bounds when $D\ge2$. Together, these results reveal collectivity as a fundamental mechanism for constraining many-body entanglement alongside spatial locality.

quant-ph↗

arXiv2026

EfficientRollout: System-Aware Self-Speculative Decoding for RL Rollouts

Reinforcement learning (RL) has become a representative post-training paradigm for large language models (LLMs), enabling strong reasoning and agentic capabilities. However, rollout generation remains a dominant latency bottleneck because autoregressive (AR) sampling decodes responses sequentially and a small number of long-tailed generations often determine completion time. Speculative decoding (SD) is a well-established technique for serving fixed LLMs that reduces latency by rapidly drafting tokens and accepting them through parallel verification while preserving the target-model distribution. However, its practical speedups do not directly carry over to RL rollouts: (i) the evolving target policy makes any fixed drafter increasingly mismatched with the policy's output distribution; and (ii) active batch sizes shrink throughout rollout decoding, shifting decoding from compute-bound to memory-bound regimes where parallel verification can exploit underutilized compute. Therefore, accelerating RL rollouts requires both a drafter that remains effective under long, high-temperature generations from an evolving policy and system-aware use of SD that avoids compute-bound regimes. We present EfficientRollout, a system-aware self-SD framework designed to address this gap for RL rollouts. EfficientRollout induces a quantized drafter from the target model (i.e. self-speculative decoding), keeping it coupled to the evolving policy without separate drafter pretraining or online adaptation. It further coordinates a system-aware SD toggle policy with acceptance-aware draft-length adaptation, enabling speculation only in beneficial regimes while matching the drafting budget to evolving drafter quality. EfficientRollout reduces rollout and end-to-end latency by up to 24.2% and 15.6%, respectively, over an accelerated AR rollout baseline, while preserving final model quality.

cs.LG↗

arXiv2026

Quantum complexity and generalized area law in fully connected models

The area law for entanglement entropy captures a fundamental constraint on the complexity of quantum many-body ground states and enables their efficient description. While the area law is rigorously established in one dimension, its status in higher-dimensional local systems remains unresolved, and it does not hold in general for geometrically non-local systems. Here, we establish a generalized area law for gapped ground states of fully connected Hamiltonians. We show that the bipartite entanglement entropy grows at most logarithmically with system size despite the absence of geometric locality. In the ground state, each site is only weakly entangled with the rest, while configurations with extensive local fluctuations are strongly suppressed, effectively restricting the accessible Hilbert space. As a consequence, the ground state admits a matrix product state approximation with polynomial bond dimension with respect to the system size at fixed accuracy. In the permutation-invariant setting, we further prove a constant entanglement bound and demonstrate that the gapped ground state can be computed in time polylogarithmic in the system size. These results show that low entanglement complexity need not rely on geometric locality, broadening the conceptual and computational scope of area laws.

quant-ph↗

arXiv2026

Entanglement area law in interacting bosons from the Bose-Hubbard model to $ϕ$4 theory and beyond

The entanglement area law is a universal principle that characterizes quantum many-body phases and underpins tensor network algorithms. Traditionally, its validity has been limited to systems with short-range interactions and bounded local energy. Achieving a complete generalization that removes both of these constraints has been a longstanding goal in quantum many-body theory, especially for interacting boson systems where unbounded energy presents intrinsic difficulties. In this work, we rigorously prove the area law for one-dimensional interacting boson systems with long-range interactions, covering broad models including the Bose-Hubbard and $\phi4$ classes. Furthermore, we establish an efficiency guarantee for Matrix-Product-State approximations of the ground states, offering a practical route to numerical simulation. One of our main technical contributions is a general method for Hilbert space dimension reduction, whose applicability extends to arbitrary spatial dimensions. These results address two major challenges simultaneously and provide important foundations for simulating long-range cold atomic systems.

quant-ph↗

arXiv2026

Optimal low-rank compression of quantum dynamics

Local quantum interactions generate dynamics in an exponentially large Hilbert space, yet locality and entanglement restrict the information that can spread and accumulate. Bounds on propagation and entanglement growth, together with tensor networks, exploit these restrictions to discard information unnecessary for describing the evolution. This leads to a sharper question: how much information must any low-rank representation retain? Here we determine these limits, up to logarithmic factors, for short-range interactions. For time-independent evolution, the optimal rank obeys $\log D=\widetilde O(t+\sqrt{\log(1/ε)})$, with matching lower bounds fixing the accuracy exponent $1/2$; for arbitrary driving, matching fixed-time bounds instead give $2/3$. Correspondingly, the corresponding dynamical entanglement spectra exhibit distinct small-$α$ Rényi laws, $α^{-1}$ and $α^{-2}$. In one dimension, the static limit is constructively attained by an explicit MPO algorithm, with an analogous extension to Liouvillian dynamics. Together, these results determine the irreducible information required to represent local quantum evolution and uncover distinct entanglement structures in static and driven dynamics.

quant-ph↗

arXiv2026

Entropy spectroscopy of a tunable two-site Hubbard molecule

Determining which microscopic states remain thermally active when tunneling and interactions compete is a central question in Hubbard physics. Here we develop an entropy measurement protocol for a double quantum dot, extending the charge-based approach established for a single dot, and measure the total entropy of a tunable two-site Hubbard molecule realized in a GaAs double dot. By shifting both dot levels together at fixed detuning, the entropy is correctly probed with only a single charge sensor. As interdot tunneling increases, the system evolves from two atomic-like dots through hybridized molecular states to a merged single dot. Across this evolution, the entropy decreases as tunnel-induced energy splitting exceeds the thermal energy and suppresses the occupation of higher-energy states. The measurements resolve the diminishing contribution of antibonding states and the changing thermal contributions of hybridized singlet and triplet states, in quantitative agreement with a two-site Hubbard model. By distinguishing states with the same charge configuration but different orbital and spin content, entropy reveals how tunneling and interactions determine the thermally active states of the minimal Hubbard system.

cond-mat.mes-hall↗

arXiv2026

Constrained Attitude Reorientation Maneuver of Spacecraft based on Path Planning

Spacecraft attitude control becomes challenging when light-sensitive instruments must avoid bright objects. This work proposes a path-planning-based spacecraft attitude reorientation maneuver considering these constraints to reach the desired attitude. Constraint-compliant attitude trajectories are computed using a fuzzy enhanced potential field approach in the attitude space. Based on the path-planning output, the control input is determined by considering the dynamics and kinematics of the spacecraft to satisfy dynamic constraints in the time domain. Numerical simulation results demonstrate successful reorientation maneuvers without violating the constraints.

physics.space-ph↗

arXiv2026

Multi-Axis Selective Decoupling Framework for Free-Floating Space Manipulators

Free-floating space manipulator systems face operational challenges due to dynamic coupling. While traditional reactionless manipulation eliminates base disturbances, enforcing the null-space projections across all rotational axes imposes excessive constraints on the system and drastically reduces the available workspace. To address this limitation, this work proposes a multi-axis selective decoupling framework that nullifies momentum transfer exclusively along mission-critical directions. By leveraging a directionally constrained sub-coupling matrix, the framework relaxes the null-space constraints. Numerical simulations confirm that this approach mathematically eliminates constrained directional disturbances along the targeted axes while preserving the remaining degrees of freedom, enabling the manipulator to execute task-space operations successfully.

cs.RO↗

arXiv2026

An Algebraic Method for Optimizing the State, Control, and Terminal State Weight Matrices for Optimal Feedback Control

The necessary conditions for formulating optimal feedback control algorithms have been known for many years. Free parameters exist in the performance index in the form of state and control penalty and terminal state penalty matrices for tuning the performance of the optimally controlled system. The selection process is typically experimental and iterative. To generate the weight matrices for optimal feedback control algorithms, an optimization process is proposed. Typically, the optimization process for the weight matrices requires several numerical integration processes that are computationally expensive; this work overcomes the classical high computational cost by exploiting closed-form solutions for the time-varying Riccati matrix, state trajectories, state transition matrix, and the optimal performance index. Closed-form algebraic equations are used to generate all partial derivative calculations, and no numerical integration is required. The closed-form partial derivatives are used to generate analytic gradients for the optimization steps. The optimization strategy seeks to minimize the terminal state values for the feedback control problem. A numerical example is presented to demonstrate the effectiveness of the proposed optimization algorithm. The resulting computational procedures are expected to be broadly useful for control theory applications in science and engineering.

math.OC↗

arXiv2026

Genetic Fuzzy System-Based Multi-Robot Coordination for Planetary Missions

This paper proposes a decentralized approach for a multi-robot system (MRS) using a genetic fuzzy system to perform a collaborative object transportation task that minimizes the total path length of the MRS in unstructured environment while avoiding obstacles. For an environment given by an elevation map, terrain traversability analysis with respect to the slope is performed to reduce the dimension and identify non-traversable areas that can be considered as obstacles, and the given map is converted into a traversability map in two dimensional space. In the training process, proposed fuzzy inference systems (FISs) to generate the MRS's velocity for transporting an object to a target position are optimized by a genetic algorithm with several scenarios, such as a local minima, a target that is close to an obstacle, and a cluttered environment. The trained FIS models are applied to the testing environment, which is the converted traversability map, and validated using multiple scenarios.

cs.RO↗

arXiv2026

Genetic Fuzzy System-based Control for Final Approach of Spacecraft Rendezvous and Proximity Operations

In-space servicing has been receiving great attention to extend the operation of spacecraft with defective components. This requires rendezvous and proximity operations for a chaser to provide service to a target. This work constructs a fuzzy inference system-based controller for the chaser to reach the cooperative target on a circular orbit in the final approach phase while minimizing the energy consumption of the chaser. The offline training process performed by a genetic algorithm deals with multiple initial relative positions of the chaser, and the trained controller is validated using a testing environment with disturbances, which differs from the training scenarios.

physics.space-ph↗

arXiv2026

A Study on Effective Initial Guess Finding Method Based on Bézier Curves: Orbit Determination Applications

In celestial mechanics, proper orbits related to missions are obtained by solving two-point boundary value problems. Since a selection method of initial value affects the convergence of the solution, developing an effective method to find an initial guess is required. In this work, Bézier curves, which can describe complicated curves and surfaces, are utilized to find the initial guess. First, the given problems are transformed into Bézier curves forms, and Bézier curves' control points, which can handle the shape of curves, are selected by solving the system of nonlinear equations. Finally, the initial guess is obtained by substituting the calculated control points to Bézier curves. To validate the performance of the proposed method, numerical simulations are conducted with respect to three kinds of orbits, which are from circular to highly elliptical orbit (HEO). The proposed method is compared to the general shooting method. The comparison results show that the initial guess calculated by Bézier curves makes finding the solution more efficient in terms of computational time and iterations. Also, it shows that the proposed method finds the solution for the HEO while the general shooting method fails to find the solution.

math.OC↗

arXiv2026

Disentangling Haldane Phase by Generalized Clifford Circuits

Disentangling transformations play a central role in the classical simulation of quantum many-body systems, yet their analytic structure and underlying mechanism remain largely unexplored. Here, we study the structure of the disentangler in the Haldane phase of spin-1 systems using generalized Clifford circuits. To this end, we extend the Clifford-circuit-augmented matrix product states (CAMPS)-based density-matrix renormalization group (DMRG) method to spin-1 systems. Within this framework, we find that the local disentanglers optimized for the Haldane phase implement the generalized Kramers--Wannier (KW) transformation, and we analytically verify its optimality for the Affleck--Kennedy--Lieb--Tasaki (AKLT) state. Beyond reducing entanglement, the KW transformation maps the Haldane phase to a phase with spontaneously broken $\mathbb{Z}_{2}$ symmetry. This mapping is distinct from the Kennedy--Tasaki transformation and provides a new unitary route from symmetry-protected topological order to symmetry breaking.

quant-ph↗