Searcharxiv⌕ Search

arXiv subjects

Yao Li

Publications and source records attributed to Yao Li.

At least 19 recordsLinked to original sources
arXiv2026

ForceLDM: A Force-aware Latent Dynamics Model for Contact-Rich Manipulation

Contact-rich manipulation requires fine-grained control and complex physical interactions with the environment, posing significant challenges for robot policy learning. Vision-only policies struggle with such tasks, as contact states and interaction forces are difficult to infer from visual observations alone. Existing methods mitigate this issue by incorporating force or tactile sensing. However, they often treat these signals as auxiliary observations without fully exploiting their role in modeling future interaction dynamics. To address this, we propose ForceLDM, a force-aware latent dynamics model that learns future contact and motion dynamics to guide action generation. Rather than directly predicting future observations, which can entangle task-relevant dynamics with irrelevant sensory details, ForceLDM learns a compact representation of future dynamics through knowledge distillation. Specifically, a teacher network with privileged future force/torque and optical flow signals extracts future dynamic features, which are then distilled into a student network trained only on current observations. This enables the student model to reason about upcoming contacts and physical dynamics without relying on future sensory inputs at deployment. To improve robustness and generalization, we further introduce a curriculum-based progressive noise injection strategy to mitigate over-reliance on future features. Experiments on five real-world contact-rich manipulation tasks demonstrate that ForceLDM outperforms state-of-the-art baselines, generalizes effectively to novel objects, and remains robust to environmental perturbations. Project page and code are available at https://tpy001.github.io/ForceLDM/.

cs.RO↗
arXiv2026

A probabilistic score estimation using the Bismut--Elworthy--Li formula

Score-learning methods generally lack pointwise supervision, which affects the training accuracy, particularly in regions where SDE trajectories are rare. To address this challenge, we introduce a novel method called \emph{BEL anchors}, which uses Monte Carlo score estimates at user-selected space-time query points. By combining a Feynman--Kac representation of the Fokker--Planck equation with the Bismut--Elworthy--Li formula, our estimator requires neither closed-form transition densities, spatial discretization, nor derivatives of the initial density. We derive its asymptotic variance and use the anchors as sparse supervised targets for existing score-learning objectives. Experiments show that BEL anchors improve implicit score matching in poorly sampled regions, prevent score-PINN collapse, and enable accurate sample regeneration in several numerical examples.

math.NA↗
arXiv2026

The Agentic Observer Hackathon: Autonomous Decision-making in Simulated Astronomical Surveys

Astronomical surveys must allocate observing time while responding to changing conditions and competing scientific priorities. We present the Agentic Observer Hackathon (AOH), a community challenge for studying autonomous survey operations. Within a synthetic multi-object survey with prescribed exposure responses and assignment constraints, decision systems select pointings, assign targets and choose exposure configurations while responding to observing requests, instrumental faults and data loss. We evaluate systems submitted by 102 teams on campaigns reserved for evaluation and analyse target completion in a reference campaign. The leading systems combine numerical exposure planning with language models for interpreting operational information and supporting longer-term decisions. Exploratory tests show improved fault reporting with model access in several systems, while revised scientific priorities produce substantial changes in ranking. The reference campaign traces target completion and diagnoses errors in the baseline's inference of prescribed quality bands. The released simulator and reference strategies provide common survey conditions for comparing exposure allocation, target selection and recovery from interruptions.

astro-ph.IM↗
arXiv2026

Continual Residual Memory for Gradient-Free Test-Time Adaptation in Speech Enhancement

Offline-trained speech enhancement models can degrade under deployment mismatch. Test-time adaptation (TTA) addresses this problem using unlabeled test observations, but existing approaches mainly rely on online parameter optimization and do not explicitly exploit the residual discarded by the enhancement backbone. We propose Continual Residual Memory (CRM), a gradient-free continual TTA framework for selective residual recovery. CRM first estimates bounded time-frequency residual recovery from the current utterance alone and then uses causal prototype memory from preceding utterances to refine this estimate. During deployment, adaptation is driven solely by the evolving memory state while the backbone and CRM modules remain frozen. Across DNS, EARS-D, and Libri-MUSAN, CRM improves all seven reported metrics over frozen CMGAN and achieves the best PESQ, CSIG, and COVL on all three test sets among the compared methods. CRM remains lightweight and shows cross-backbone applicability on five additional backbones. Code is available at https://github.com/Ada312/continual-residual-memory.

eess.AS↗
arXiv2026

Nonreciprocity Propels Nematic Defects

Nonreciprocal interactions are a generic source of nonequilibrium dynamics, yet their consequences for nematic topological defects remain unexplored. We introduce a nonreciprocal nematic model on a two-dimensional hexagonal lattice and formulate a continuum theory containing a nematic nonreciprocal torque. In the deterministic limit, simulations and symmetry analysis show that this torque drives +1/2 defects to self-propel along their polar axes. Nonreciprocity modifies the splitting pathway of integer +1 defects. These results identify a symmetry-resolved route from nonreciprocity to nematic defect transport and topological reorganization, distinct from active-stress-driven motion.

cond-mat.soft↗
arXiv2026

A Deep Learning-Driven Operator Splitting Scheme for Free Boundary Problems: An Application to American Option Pricing in Incomplete Markets

This paper proposes a deep learning-driven operator splitting scheme for high-dimensional free-boundary problems, with American option pricing as a representative application. In a market with proportional transaction costs, a stochastic liquidity risk with an empirically validated nonlinear drift based on S&P 500 data, and two-factor stochastic volatility, the pricing problem takes the form of a five-dimensional highly nonlinear variational inequality. The proposed scheme decomposes this variational inequality at each time layer into an elliptic continuation equation approximated by a physics-informed neural network and a closed-form projection step enforcing the early-exercise constraint. A conditional $\ell^\infty$ stability bound is derived under stated resolvent and Lipschitz assumptions. The scheme is validated against finite-difference and least-squares Monte Carlo benchmarks, and both the option prices and the learned exercise boundary show close agreement with the reference solutions. Sensitivity analyses with respect to the key model parameters are also reported.

q-fin.MF↗
arXiv2026

A Taxonomy of Predicate Transformer Monads for Specifying Monadic Programs

Monads of weakest precondition predicate transformers are the foundation of several verification frameworks for monadic programs. For specifying monadic computations with nondeterminism and nontermination, in addition to weakest preconditions, we introduce the first strongest postcondition monads covering all combinations of angelic or demonic nondeterminism and partial or total interpretations of nontermination. For extending these base strongest postcondition monads with side effects such as mutable state, monad transformers do not yield the expected monads, so we instead generalize the base monads to arrows and use arrow transformers. We give a taxonomy of weakest precondition and strongest postcondition monads, for the base monads and for their versions with state, and show that the known relationships between under- and over-approximate triples based on predicate transformers already hold at the level of the predicate transformers themselves. We use the predicate transformer monads to define under- and over-approximate triples and, where possible, Dijkstra monads. Finally, we relate these two specification styles.

cs.PL↗
arXiv2026

From Clicks to Intent: Cross-Platform Session Embeddings with LLM-Distilled Taxonomy for Financial Services Recommendations

Sequential user behavior modeling is widely adopted in industrial recommender systems; however, significant gaps remain in financial services, where pre-login web interactions and authenticated in-app experiences differ drastically. Specifically, pre-login web users typically explore new products, whereas logged-in app users focus on account servicing. Due to the challenge of cross-channel entity resolution (e.g., matching anonymous web sessions to authenticated mobile accounts), web-based intent signals remain underutilized for post-authentication personalization. Existing methods for capturing web-based intent are often ad-hoc and narrow, lacking the flexibility to support both quantitative downstream recommendations and qualitative understanding at scale. In this work, we propose a scalable and dual-purpose intent prediction framework for web-based interactions and demonstrate its applicability for personalization. Our approach transforms raw web clickstreams into two outputs: a self-supervised Transformer encodes multi-modal clickstreams into a compact session embedding, while an LLM-based taxonomy generation and distillation pipeline produces interpretable intent labels. Our system demonstrates that self-supervised clickstream representations combined with LLM-distilled taxonomies can jointly serve quantitative tasks and qualitative understanding in production: on the mobile homepage tile ranking task, the session embedding improves macro Recall@1 by 1.88% and reduces Log Loss by 13.38% over production baselines. On the user conversion prediction task, the embedding outperforms the LLM labels by 4.3% on micro F1, while the distillation layer delivers interpretable labels at ultra-low latency with only a 7% performance drop.

cs.IR↗
arXiv2026

UMR: Universal Manipulation Representation

General-purpose embodied manipulation hinges on a unified action representation that generalizes across embodiments and scales readily. Yet existing policies rely on embodiment-specific action spaces, making cross-embodiment demonstrations difficult to leverage at scale and limiting transfer to new embodiments and spatial variations. To this end, we introduce Universal Manipulation Representation (UMR), a unified action representation that enables zero-shot skill transfer from human demonstrations to heterogeneous robots. UMR decomposes manipulation into two functionally distinct yet geometrically linked components: embodiment-agnostic World Flow, which describes task-relevant object motion in the world frame, and Ego Trajectory, which represents end-effector motion relative to the current pose. We instantiate UMR as World--Ego Point VLA (WEPVLA), a compact 0.5B-parameter policy that learns in the unified geometric action space through a dual-stream Point Action Adapter and a unified Point Action Expert, with an $SE(3)$ conjugation coupling the two components. To improve data efficiency, we complement UMR with a Data-Efficient Strategy (DES) that diversifies object configurations through stage-aware point-cloud editing while preserving demonstrated contact geometry. In simulation, WEPVLA achieves average success rates of 97.5\% on LIBERO and 85.7\% on the 10-task RLBench benchmark. In real-world experiments, a single policy trained on human demonstrations augmented by DES transfers zero-shot to diverse deployment conditions. With about 10 minutes of collected human demonstrations per task and no robot demonstrations, it achieves 91.7\% average success across six evaluation settings, compared with 60.8\% for HumanEgo. Code and additional materials are available at https://umr-wepvla.github.io/.

cs.RO↗
arXiv2026

Series-parallel extensions and Ehrhart positivity

We prove Ehrhart positivity for series-parallel matroids and extend this result to all series-parallel extensions of multi-path matroids. A formula for the Ehrhart coefficients of series-parallel matroids is derived and used to prove strict positivity. An extension criterion based on weighted lattice-point counts is established. Recursive formulas for lattice-point counts are derived, and the recursion is shown to preserve a class of polynomials with nonnegative coefficients. We finally construct a connected transversal matroid with a negative Ehrhart coefficient, disproving the conjecture of Ferroni, Morales, and Panova that every connected matroid whose base polytope admits a subdivision into series-parallel matroid base polytopes is Ehrhart positive.

math.CO↗
arXiv2026

Defect Organization in Coexisting Hexagonal and Square Lattices on Ellipsoids

Curvature and topology jointly organize defects in two-dimensional crystals, but their combined role remains unresolved when competing lattice symmetries coexist with spatially varying curvature. We use simulated-annealing Langevin dynamics to study Hertzian particles forming coexisting hexagonal (Hex) and square (Sq) lattices on prolate and oblate ellipsoids. Mapping reduced density and aspect ratio reveals a broad sequence of scar and domain-based morphologies in both Hex-dominant and Sq-dominant backgrounds. Latitude-resolved comparisons show that Gaussian curvature biases defects toward its maxima under weak deformation. Strong prolateness, however, confines high curvature to small polar caps that cannot independently accommodate all defect motifs. Defects then spread toward lower-curvature latitudes to relieve defect crowding and elastic repulsion. In the Hex-dominant regime, this competition drives vertex-contacted domains with neutralized corner contacts, and compensating positive defects locate away from the poles. The Sq-dominant regime features Hex-rich triangular domains, bridged states, and linear or open scars similarly reorganized by curvature anisotropy. On oblate ellipsoids, the extended equatorial high-curvature belt allows defects to separate azimuthally while remaining curvature-localized. This work elucidates that nonuniform curvature can engineer rich defect patterns by selecting the spatial distribution of topological charge and the connectivity of finite defect motifs.

cond-mat.soft↗
arXiv2026

Achieving 5-angstrom-resolution diffraction contrast with an uncorrected electron probe and nanosized defect characterizations

Atomic-resolution micrographs remain limited in application due to small field of view and low numbers of recorded defects. Defect imaging with reliable statistics in transmission electron microscopy (TEM) relies heavily on diffraction contrast. However, with 200-300 kV electrons, low-curvature Ewald sphere limits diffraction resolution at ~5 nm. Here, we imaged defects with a forbidden condition in TEM and improved the resolution in diffraction contrast from 5 to 0.5 nm. The superiority of our equal-s method is demonstrated by the discovery of a novel phase and dislocation loop structure in steels, which have not been previously reported experimentally or theoretically. Our equal-s can largely replace the classical two-beam condition for defect study.

cond-mat.mtrl-sci↗
arXiv2026

A Tunable Despeckling Neural Network Stabilized via Diffusion Equation

The removal of multiplicative Gamma noise is a critical research area in the application of synthetic aperture radar (SAR) imaging, where neural networks serve as a potent tool. However, real-world data often diverges from theoretical models, exhibiting various disturbances, which makes the neural network less effective. Adversarial attacks can be used as a criterion for judging the adaptability of neural networks to real data, since they can find the most extreme perturbations that make neural networks ineffective. In this work, we propose a tunable, regularized neural network framework that unrolls a shallow neural denoising block and a diffusion regularization block into a single network for end-to-end training. The linear heat equation, known for its inherent smoothness and low-pass filtering properties, is adopted as the diffusion regularization block. The smoothness of our outputs is controlled by a single time step hyperparameter that can be adjusted dynamically. The stability and convergence of our model are theoretically proven. Experimental results demonstrate that the proposed model effectively eliminates high-frequency oscillations induced by adversarial attacks. Finally, the proposed model is benchmarked against several state-of-the-art denoising methods on simulated images, adversarial samples, and real SAR images, achieving superior performance in both quantitative and visual evaluations.

cs.CV↗
arXiv2026

A General Multimodal Probing Framework for Data Fusion and Prediction with Frozen Large Language Models: Evidence from Multimodal EHR Data

The main objective of this paper is to propose a general framework for prediction based on different sources of multimodal data in the healthcare domain. We evaluated whether frozen medical large language model (LLM) representations can serve as a shared embedding space for multimodal primary diagnosis category prediction. We propose a backbone-flexible probing framework that serializes structured electronic health record (EHR) variables and leakage-pruned discharge-note sections into a shared frozen-LLM representation space and recovers primary diagnosis categories with linear probes, without fine-tuning the LLM. On a MIMIC-IV cohort of 13,645 admissions (seven diagnosis categories), we probed inputs of different modalities (Structured-only, Unstructured-only, and Combined) across transformer layers and six frozen backbones, compared against XGBoost and three text baselines (CAML, LAAT, PLM-ICD) under a matched split and context budget, and tested robustness across datasets on MIMIC-III. The Combined terminal-layer probe reached 88.86% top-1 accuracy on MIMIC-IV, significantly outperforming CAML and PLM-ICD and statistically comparable to LAAT; Combined accuracy ranged from 87.69% to 89.25% across six backbones. On MIMIC-III, in-domain Combined probing reached 74.18% accuracy and significantly exceeded all three baselines, and in a transfer-learning test a 2.8M-parameter adapter on frozen MIMIC-IV representations reached 92.51% accuracy, exceeding adapted LAAT (85.54%). Frozen LLM embeddings unify structured and narrative EHR information: the Combined configuration was strongest in every setting, and its representations transferred across datasets through a compact adapter. Multimodal probing of frozen LLM representations provides a practical, backbone-flexible approach for studying EHR modalities and adapting clinical representations across datasets.

cs.AI↗
arXiv2026

QART: A Quantum-Classical Hybrid Architecture for Long-Horizon Reasoning -- Exploring a Conditional Path toward Quantum Scaling

Long-horizon reasoning is vulnerable to early errors that compromise later decisions. We present QART, the Quantum-Augmented Reasoning Transformer, a quantum--classical hybrid architecture combining a backbone language model with quantum encoding, CIM-based QUBO optimization, and quantum decoding. Semantic information can come from hidden representations or model-generated text; detailed encoding and optimization procedures remain proprietary. Under explicit assumptions, we establish a conditional asymptotic reliability separation from single-trajectory autoregressive LLMs. For a common task family with aligned optimality and acceptance criteria, autoregressive acceptance probability tends to zero when cumulative conditional risk of irreversible errors diverges. QART's task-optimal-path recovery probability remains bounded away from zero if conditional probabilities for optimal-path coverage and semantic fidelity, spectral certification, dynamical reachability, and faithful readout remain uniformly positive under a specified resource schedule. The architecture alone does not imply these bounds. Paired measurements on six long-horizon benchmarks using DeepSeek V4 Flash, GLM-5.3, and GPT-5.5 xhigh in a Codex agent environment favor QART in 14 of 15 backbone--benchmark pairs. Relative gains reach 84.0% on SciCode, 47.6% on $τ^3$-Bench, and 44.4% on Terminal-Bench 4.0; the DeepSeek V4 Flash configuration regresses by 7.8% on DeepSWE. These results do not directly validate the asymptotic separation. Potential quantum scaling laws are formulated as conditional hypotheses. A quantum-advantage interpretation requires a demonstrated CIM quantum advantage over strong classical solvers and its transfer to end-to-end reasoning after all system overheads.

cs.AI↗
arXiv2026

SAR-FAH: A Frequency-Adaptive Hybrid Network based on Neural ODEs for Structural-Preserving SAR Despeckling

Synthetic Aperture Radar (SAR) images are inherently degraded by speckle noise that severely limits their reliability in high-precision applications. As a signal-dependent multiplicative noise, speckle noise exhibits distinct statistical properties in homogeneous and heterogeneous regions of SAR images, which are spatially coupled. Nevertheless, existing deep learning despeckling methods operate directly in the spatial domain overlooking this statistical difference. It imposes a suboptimal trade-off between noise suppression and structure preservation, inevitably leading to artifacts, edge blurring, and texture distortion. To address these limitations, we propose a Frequency-Adaptive Hybrid model based on Neural Ordinary Differential Equations (NODEs) for SAR despeckling, termed SAR-FAH. It is a novel divide-and-conquer architecture that performs despeckling in the frequency domain to achieve improved structural preservation. We first fully decouple homogeneous and heterogeneous regions in the frequency domain via wavelet transform according to their local spatial characteristics and then revisit the statistical characteristics of speckle noise. Guided by the distinct properties of each sub-band, we design specialized sub-networks for frequency-specific restoration. Specifically, based on the smoothness of the low-frequency sub-band, the low-frequency denoising process is controlled by the module based on NODEs to ensure sufficient smoothness without artifacts, while the high-frequency sub-bands are processed by enhanced U-Net by incorporating deformable convolutions to better suppress noise and preserve edges and textures. Extensive experiments on both synthetic and real SAR images demonstrate that the proposed SAR-FAH outperforms the state-of-the-art methods both quantitatively and qualitatively.

cs.CV↗
arXiv2026

On the Frankl--Tokushige conjecture and almost complete $r$-cross $t$-intersection theorems for vector spaces

Let $r\geq3$ and $k_1\geq k_2\geq\cdots\geq k_r\geq t$. Let $\mathcal{F}_1,\mathcal{F}_2,\ldots,\mathcal{F}_r$ be families of subspaces, of respective dimensions $k_1,k_2,\ldots,k_r$, in an $n$-dimensional vector space over the finite field $\mathbb{F}_q$. The $r$ families are called $r$-cross $t$-intersecting if $\dim \left(F_{1} \cap F_{2} \cap \cdots \cap F_{r}\right) \geq t$ for all $F_{i} \in \mathcal{F}_{i}, i = 1,2,\dots,r$. In 2016, Frankl and Tokushige conjectured that $\prod_{i=1}^{r}|\mathcal{F}_i|\leq\prod_{i=1}^{r}{n-1\brack k_i-1}$ for $t=1$ and $n\geq rk_1/(r-1)$. The appealing conjecture suggests establishing intersection theorems for $n\sim ck_1$ with $c=c(r)\in(1,2)$, a direction that has long been challenging. In this paper, we overcome this barrier by proving that $$\prod_{i=1}^{r}|\mathcal{F}_i|\leq\prod_{i=1}^{r}{n-t\brack k_i-t}\;\;\mbox{for all}\;\;t\geq1\;\mbox{and}\;n\geq rk_1/(r-1)+C(t,r),$$ where $C(t,r)=rt/(r-1)+1$. This proves the Frankl--Tokushige conjecture except for at most three values of $n$, and establishes an Erdős--Ko--Rado type theorem for almost all values of parameters. Furthermore, we characterize all extremal configurations. Our proof is purely combinatorial and based on the $t$-cover method, with several essential refinements. We also obtain almost complete intersection theorems for $r$-wise $t$-intersecting families and non-trivial $r$-cross $t$-intersecting families.

math.CO↗
arXiv2026

The Late-time Ramp of the Double-Scaled SYK Model from the large-$n$ tail of Cactus Diagram

We derive the late-time ramp of the finite-temperature spectral form factor in the double-scaled SYK model directly from cactus diagrams, which is introduced as a multi-trace generalization of chord diagram in Ref \cite{Berkooz:2020fvm}. Although single-trace observables admit an exact chord-diagram description, multi-trace sums are obstructed by the chord-intersection weights $q_{IJ}$, which cannot be averaged independently to $q$. Our key idea is that the non-analytic contribution responsible for the ramp is controlled by the large-order tail of the Cactus-diagram expansion and is insensitive to finite changes in its low-order analytic terms. Decomposing the contribution with $n$-cross-trace-pairings Cactus diagram into two buds $B_n$ and a kernel $\mathcal K_n$, we determine their large-$n$ asymptotics with fixed $0\leq q<1$ and resum the resulting tail. For $β_L=β+it$ and $β_R=β-it$, we obtain \begin{equation*} Z_s^{\mathrm{sing}}(β+it,β-it) =s_p c_N \frac{|t|}{2π} \int_{E_{min}}^{E_{max}} e^{-2βE} dE+O(1), \qquad s_p=\begin{cases}2,&4\mid p,\\1,&4\nmid p.\end{cases} \end{equation*} This result reproduces the linear ramp predicted by random matrix theory, including its temperature dependence and symmetry factor, and agrees with the semiclassical predictions. It provides a direct microscopic origin of the ramp within the general $q$-deformed quantum algebra. While the plateau lies beyond the scope of the present analysis and calculation.

hep-th↗