arXiv subjects

Jiefu Zhang

Publications and source records attributed to Jiefu Zhang.

10 recordsLinked to original sources

arXiv2026

Selection-Aware Stress Testing for Interactive Agents

Agent evaluations often use one benchmark to choose a workflow and then search for task types where its advantage weakens, so both conclusions are selected from the same data. We introduce Selection-Aware Semantic Stress Testing (\SASST{}), which learns a task reweighting from pre-execution features on discovery tasks and evaluates the same paired comparison on separate confirmation tasks. The protocol checks support and stability, uses joint bounds for all planned claims, and can return no claim. We prove conditional asymptotic validity under stated cluster assumptions. A forty-cluster audit finds Gaussian undercoverage and conservative Bonferroni $t$ bounds. In one 480-episode $τ$-bench study, a $3.75$ point discovery gain vanished on confirmation. A second-model study likewise confirmed neither a workflow benefit nor a stable stress rule.

cs.LG↗

arXiv2026

Versatile Exoskeleton Control Supports Elderly Joint Energetics During Hip-Intensive Tasks

Age-related mobility decline is frequently accompanied by a redistribution of joint kinetics, where older adults compensate for reduced ankle function by increasing demand on the hip. Paradoxically, this compensatory shift typically coincides with age-related reductions in maximal hip power. Although robotic exoskeletons can provide immediate energetic benefits, conventional control strategies have limited previous studies in this population to specific tasks such as steady-state walking, which do not fully reflect mobility demands in the home and community. Here, we implement a task-agnostic hip exoskeleton controller that is inherently sensitive to joint power and validate its efficacy in eight older adults. Across a battery of hip-intensive activities that included level walking, ramp ascent, stair climbing, and sit-to-stand transitions, the exoskeleton matched biological power profiles with high accuracy (mean cosine similarity 0.89). Assistance significantly reduced sagittal plane biological positive work by 24.7\% at the hip and by 9.3\% for the lower limb, while simultaneously augmenting peak total (biological + exoskeleton) hip power and reducing peak biological hip power and torque. These results suggest that hip exoskeletons can potentially enhance endurance through biological work reduction, and increase functional reserve through total power augmentation, serving as a promising biomechanical intervention to support elderly mobility.

cs.RO↗

arXiv2026

Core-Halo Decomposition: Decentralizing Large-Scale Fixed-Point Problems

We study solving large-scale fixed-point equation $x^\star=\bar F(x^\star)$ with decomposition. Standard strict decomposition assigns each agent a disjoint block and evaluates updates using only owned coordinates. For most operators, however, a block update may depend on variables outside the block. Truncating these dependencies by strict decomposition changes the mean operator and creates structural bias that cannot be removed by more samples, smaller stepsizes, or additional consensus. We therefore propose Core-Halo decomposition, which separates write ownership from read-only evaluation context: each agent updates its own core and reads from an overlapping halo. By aligning the Core-Halo decomposition with the block-dependence structure of $\bar F$, the original fixed-point problem can be implemented faithfully in a decentralized multi-agent system. We further characterize the fundamental obstruction faced by strict decomposition through a Bellman closure condition and a blockwise bias lower bound, showing that local-only updates can alter the original fixed-point operator. Finally, we conduct extensive experiments across a range of application settings, and demonstrate that Core-Halo achieves near-centralized performance while retaining the parallelism benefits of decentralization.

stat.ML↗

arXiv2026

Towards Reliable LLM Evaluation: Correcting the Winner's Curse in Adaptive Benchmarking

Adaptive prompt and program search makes LLM evaluation selection-sensitive. Once benchmark items are reused inside tuning, the observed winner's score need not estimate the fresh-data performance of the full tune-then-deploy procedure. We study inference for this procedure-level target under explicit tuning budgets. We propose SIREN, a selection-aware repeated-split reporting protocol that freezes the post-search shortlist, separates splitwise selection from held-out evaluation, and uses an item-level Gaussian multiplier bootstrap for uncertainty quantification. In a fixed-shortlist regime with smooth stabilized selection, the estimator admits a first-order item-level representation, and the bootstrap yields valid simultaneous inference on a finite budget grid. This supports confidence intervals for procedure-performance curves and pre-specified equal-budget and cross-budget comparisons. Controlled simulations and MMLU-Pro tuning experiments show that winner-based reporting can be optimistic and can change deployment conclusions, while SIREN remains close to the finite-sample reporting target.

stat.ML↗

arXiv2026

Don't Freeze, Don't Crash: Extending the Safe Operating Range of Neural Navigation in Dense Crowds

Navigating safely through dense crowds requires collision avoidance that generalizes beyond the densities seen during training. Learning-based crowd navigation can break under out-of-distribution crowd sizes due to density-sensitive observation normalization and social-cost scaling, while analytical solvers often remain safe but freeze in tight interactions. We propose a reinforcement learning approach for dense, variable-density navigation that attains zero-shot density generalization using a density-invariant observation encoding with density-randomized training and physics-informed proxemic reward shaping with density-adaptive scaling. The encoding represents the distance-sorted $K$ nearest pedestrians plus bounded crowd summaries, keeping input statistics stable as crowd size grows. Trained with $N\!\in\![11,16]$ pedestrians in a $3\mathrm{m}\times3\mathrm{m}$ arena and evaluated up to $N\!=\!21$ pedestrians ($1.3\times$ denser), our policy reaches the goal in $>99\%$ of episodes and achieves $86\%$ collision-free success in random crowds, with markedly less freezing than analytical methods and a $>\!60$-point collision-free margin over learning-based benchmark methods. Codes are available at \href{https://github.com/jznmsl/PSS-Social}{https://github.com/jznmsl/PSS-Social}.

cs.LG↗

arXiv2025

Rapid MRI-Based Synthetic CT Simulations for Precise tFUS Targeting

Accurate targeting is critical for the effectiveness of transcranial focused ultrasound (tFUS) neuromodulation. While CT provides accurate skull acoustic properties, its ionizing radiation and poor soft tissue contrast limit clinical applicability. In contrast, MRI offers superior neuroanatomical visualization without radiation exposure but lacks skull property mapping. This study proposes a novel, fully CT free simulation framework that integrates MRI-derived synthetic CT (sCT) with efficient modeling techniques for rapid and precise tFUS targeting. We trained a deep-learning model to generate sCT from T1-weighted MRI and integrated it with both full-wave (k-Wave) and accelerated simulation methods, hybrid angular spectrum (kWASM) and Rayleigh-Sommerfeld ASM (RSASM). Across five skull models, both full-wave and hybrid pipelines using sCT demonstrated sub-millimeter targeting deviation, focal shape consistency (FWHM ~3.3-3.8 mm), and <0.2 normalized pressure error compared to CT-based gold standard. Notably, the kW-ASM and RS-ASM pipelines reduced simulation time from ~3320 s to 187 s and 34 s respectively, achieving ~94% and ~90% time savings. These results confirm that MRI-derived sCT combined with innovative rapid simulation techniques enables fast, accurate, and radiation-free tFUS planning, supporting its feasibility for scalable clinical applications.

physics.med-ph↗

arXiv2023

Optimal Energy Shaping Control for a Backdrivable Hip Exoskeleton

Task-dependent controllers widely used in exoskeletons track predefined trajectories, which overly constrain the volitional motion of individuals with remnant voluntary mobility. Energy shaping, on the other hand, provides task-invariant assistance by altering the human body's dynamic characteristics in the closed loop. While human-exoskeleton systems are often modeled using Euler-Lagrange equations, in our previous work we modeled the system as a port-controlled-Hamiltonian system, and a task-invariant controller was designed for a knee-ankle exoskeleton using interconnection-damping assignment passivity-based control. In this paper, we extend this framework to design a controller for a backdrivable hip exoskeleton to assist multiple tasks. A set of basis functions that contains information of kinematics is selected and corresponding coefficients are optimized, which allows the controller to provide torque that fits normative human torque for different activities of daily life. Human-subject experiments with two able-bodied subjects demonstrated the controller's capability to reduce muscle effort across different tasks.

cs.RO↗

arXiv2022

Universal approximation of symmetric and anti-symmetric functions

We consider universal approximations of symmetric and anti-symmetric functions, which are important for applications in quantum physics, as well as other scientific and engineering computations. We give constructive approximations with explicit bounds on the number of parameters with respect to the dimension and the target accuracy $ε$. While the approximation still suffers from the curse of dimensionality, to the best of our knowledge, these are the first results in the literature with explicit error bounds for functions with symmetry or anti-symmetry constraints.

math.NA↗

arXiv2020

Learning the mapping $\mathbf{x}\mapsto \sum_{i=1}^d x_i^2$: the cost of finding the needle in a haystack

The task of using machine learning to approximate the mapping $\mathbf{x}\mapsto\sum_{i=1}^d x_i^2$ with $x_i\in[-1,1]$ seems to be a trivial one. Given the knowledge of the separable structure of the function, one can design a sparse network to represent the function very accurately, or even exactly. When such structural information is not available, and we may only use a dense neural network, the optimization procedure to find the sparse network embedded in the dense network is similar to finding the needle in a haystack, using a given number of samples of the function. We demonstrate that the cost (measured by sample complexity) of finding the needle is directly related to the Barron norm of the function. While only a small number of samples is needed to train a sparse network, the dense network trained with the same number of samples exhibits large test loss and a large generalization gap. In order to control the size of the generalization gap, we find that the use of explicit regularization becomes increasingly more important as $d$ increases. The numerically observed sample complexity with explicit regularization scales as $\mathcal{O}(d^{2.5})$, which is in fact better than the theoretically predicted sample complexity that scales as $\mathcal{O}(d^{4})$. Without explicit regularization (also called implicit regularization), the numerically observed sample complexity is significantly higher and is close to $\mathcal{O}(d^{4.5})$.

cs.LG↗

arXiv2019

Deep Density: circumventing the Kohn-Sham equations via symmetry preserving neural networks

The recently developed Deep Potential [Phys. Rev. Lett. 120, 143001, 2018] is a powerful method to represent general inter-atomic potentials using deep neural networks. The success of Deep Potential rests on the proper treatment of locality and symmetry properties of each component of the network. In this paper, we leverage its network structure to effectively represent the mapping from the atomic configuration to the electron density in Kohn-Sham density function theory (KS-DFT). By directly targeting at the self-consistent electron density, we demonstrate that the adapted network architecture, called the Deep Density, can effectively represent the electron density as the linear combination of contributions from many local clusters. The network is constructed to satisfy the translation, rotation, and permutation symmetries, and is designed to be transferable to different system sizes. We demonstrate that using a relatively small number of training snapshots, Deep Density achieves excellent performance for one-dimensional insulating and metallic systems, as well as systems with mixed insulating and metallic characters. We also demonstrate its performance for real three-dimensional systems, including small organic molecules, as well as extended systems such as water (up to $512$ molecules) and aluminum (up to $256$ atoms).

physics.comp-ph↗