Searcharxiv⌕ Search

arXiv subjects

Wei Li

Publications and source records attributed to Wei Li.

At least 109 records · Page 6Linked to original sources
arXiv2026

Learn to Memorize: Scalable Continual Learning in Semiparametric Models with Mixture-of-Neighbors Induction Memory

Semiparametric language models (LMs) have shown promise in various Natural Language Processing (NLP) tasks. However, they utilize non-parametric memory as static storage, which lacks learning capability and remains disconnected from the internal information flow of the parametric models, limiting scalability and efficiency. Based on recent interpretability theories of LMs, we reconceptualize the non-parametric memory represented by $k$NN-LM as a learnable Mixture-of-Neighbors Induction Memory (MoNIM), which synergizes the induction capabilities of attention heads with the memorization strength of feed-forward networks (FFN). By integrating into the model's information flow, MoNIM functions as an FFN-like bypass layer within the Transformer architecture, enabling effective learning of new knowledge. Extensive experiments demonstrate that MoNIM is a retentive and scalable continual learner in both data- and model-wise, enhancing the scalability and continual learning performance of semiparametric LMs.

cs.CL↗
arXiv2026

Representation-Aligned Tactile Grounding for Contact-Rich Robotic Manipulation

Tactile-enhanced vision-language-action (VLA) policies have been introduced for contact-rich manipulation, where critical interaction states are often hidden from vision. Future tactile prediction is a promising way to use touch because it turns tactile outcomes into supervision for action-induced contact dynamics. Yet VLA policies contain representations with different roles, from perceptual encoding to motor prediction, making it unclear where this supervision should be applied. We study this as a representation-alignment problem. Through a linear probe analysis, we find that future tactile states are most predictable from intermediate action-expert features, rather than from vision-language features or final action states. Motivated by this observation, we introduce a lightweight Latent Tactile Predictor (LTP), which predicts compact future tactile embeddings from the identified intermediate representation. By avoiding direct prediction of noisy raw tactile signals, LTP provides an action-outcome grounding signal that aligns intermediate action representations with future contact consequences. Experiments on real-world contact-rich manipulation tasks show that representation-aligned tactile grounding outperforms less aligned or multi-interface tactile prediction, highlighting the importance of where tactile supervision is applied.

cs.RO↗
arXiv2026

FoMoVLA: Bridging Visual Foresight and Motion Guidance for Vision-Language-Action Models

Vision-Language-Action (VLA) models have achieved impressive results in visuomotor policy learning, yet remain fundamentally reactive, mapping current observations and language to actions without explicit forward prediction of world dynamics. Existing visual foresight methods predict future visual states but lack explicit motion guidance: they show where to go but not how to get there. We argue that future feature prediction and sparse point tracking are naturally complementary: the former provides the goal state, while the latter captures the continuous motion path toward it. We propose FoMoVLA, a framework that augments VLA representations with explicit spatio-temporal supervision by jointly learning future feature foresight and sparse 2D point tracking, enhancing the continuous action policy. FoMoVLA introduces compact foresight tokens to decode future feature states, decodes sparse temporal 2D point trajectories to model compact geometric motion, and couples both through a lightweight future-conditioned cross-attention module that enables consistent reasoning between anticipated states and point dynamics. Extensive experiments on LIBERO, RoboCasa GR-1 Tabletop, and LIBERO-Plus demonstrate state-of-the-art performance and strong zero-shot generalization. Project page is available at https://liauto-research.github.io/FoMoVLA.

cs.CV↗
arXiv2026

SIVA-RL: Sensitivity-Invariance Visual Alignment for Multimodal Reinforcement Learning

Reinforcement learning with verifiable rewards (RLVR) drives multimodal reasoning, but answer-level correctness does not guarantee that a vision-language model grounds its predictions in visual evidence. Existing visual-intervention methods contrast policy behavior on original and modified images, yet assign supervision by the type of intervention rather than its observed effect. This assumption fails: identical operators produce heterogeneous outcomes across samples. We propose SIVA-RL, a Sensitivity-Invariance Visual Alignment framework that replaces operator-conditioned regularization with sample-wise, outcome-conditioned supervision. SIVA-RL constructs localized interventions through token-aligned, distance-constrained within-image PatchSwap. A frozen audit policy then scores each clean-intervention pair, and the observed reward drop becomes soft routing weights. Large-drop pairs drive sensitivity alignment, low-drop pairs drive clean-anchored invariance alignment, and ambiguous pairs are down-weighted. This design decouples intervention construction from supervision assignment and is compatible with both GRPO and DAPO backbones. Across nine multimodal reasoning benchmarks spanning mathematical, logical, and vision-dependent tasks, SIVA-RL improves 3B and 7B models over matched RL baselines in every setting. It yields an 8.79 percentage-point gain on vision-dependent reasoning and up to 14.9% relative overall improvement across all four GRPO- and DAPO-based configurations.

cs.CV↗
arXiv2026

Dissipationless dynamics of spin supersolid states in a spin-1/2 triangular antiferromagnet with impurities

Motivated by recent experimental evidence for spin supersolid states in triangular-lattice compounds, we numerically investigate the dynamical properties of magnetic field-induced phases in the spin-1/2 easy-axis triangular antiferromagnetic Heisenberg model in the presence of magnetic impurities. In both weak- and strong-field spin supersolid states, the gapless Goldstone mode at the $K$ points remains robust against impurities, which is a direct manifestation of spin superfluidity. By contrast, at the same impurity density, impurities induce a splitting of the magnon bands in the conventional magnetic state, the so-called up-up-down state. In addition, the finite superfluid stiffness probed by the twisted phase in the spin supersolid states is consistent with the excitation spectrum. We argue that the excitation spectrum with impurities provides direct spectroscopic evidence for dissipationless spin dynamics in the spin supersolid states, which is experimentally accessible via inelastic neutron scattering.

cond-mat.str-el↗
arXiv2026

DeepTravel: An End-to-End Agentic Reinforcement Learning Framework for Autonomous Travel Planning Agents

Travel planning (TP) agent has recently worked as an emerging building block to interact with external tools/resources for travel itinerary generation, ensuring an enjoyable user experience. Despite its benefits, existing studies rely on hand-craft prompt and fixed agent workflow, hindering more flexible and autonomous TP agents. This paper proposes DeepTravel, an end-to-end agentic reinforcement learning framework for building an autonomous travel planning agent, capable of autonomously planning, executing tools, and reflecting on tool responses to explore, verify, and refine intermediate actions in multi-step reasoning. To achieve this, we first construct a robust travel sandbox by caching transportation, accommodation and POI data, facilitating TP agent training without being constrained by real-world APIs limitations (e.g., inconsistent outputs). Moreover, we develop a hierarchical reward modeling system, where a trajectory-level verifier first checks spatiotemporal feasibility and filters unsatisfied travel itinerary, and then the turn-level verifier further validate itinerary's detail consistency with tool responses, enabling efficient and precise reward service. Finally, we propose the reply-augmented reinforcement learning method that enables TP agent to periodically replay from a failure experience buffer, emerging notable agentic capacity. We deploy the trained TP agent in the DiDi Enterprise Solutions application. A three-month online test shows that it achieves 82% accuracy in travel itinerary generation. Comprehensive offline evaluations further demonstrate that DeepTravel enables small-sized LLMs (e.g., Qwen3-32B) to significantly outperform frontier LLMs (e.g., OpenAI o1/o3 and DeepSeek-R1) and existing TP agent frameworks.

cs.AI↗
arXiv2026

AutoMem: A Text-Gradient Recursive Self-Improvement Framework for Automated Memory Architectures Search

Long-term memory is increasingly central to LLM agents, yet memory design remains a highly coupled architecture problem: what to encode, how to store it, how to retrieve it, and how to manage it can vary substantially across tasks and backbone models. We construct a discrete search space with 5 encoders, 5 stores, 6 retrievers, and 4 managers, and show that no single memory architecture consistently dominates: different tasks favor different module combinations, leading to substantial performance gaps. Motivated by this, we propose \textsc{AutoMem}, a text-gradient recursive self-improvement framework for task-adaptive memory architecture search. \textsc{AutoMem} optimizes over the factored space through two components: Experience-Guided Architecture Search, which proposes candidate architectures from historical search trajectories and accumulated reflections, and Failure-Guided Module Diagnosis, which localizes memory-related failures to specific modules and converts them into targeted textual feedback. Experiments on GAIA, WebWalkerQA, and xBench-DeepSearch across two LLM backbones show that \textsc{AutoMem} consistently discovers task-adaptive memory architectures that outperform the strongest human-designed memory baselines, improving accuracy by $2.8$ points on average across six benchmark-backbone settings. Further analysis shows that \textsc{AutoMem} achieves a favorable accuracy-efficiency trade-off, reducing token cost by $14.3\%$ over the strongest accuracy baselines under Qwen3.5-122B-A10B, while also finding stronger architectures than substantially larger random searches within only a few guided iterations.

cs.CL↗
arXiv2026

Active control of the peak value of the Hanbury Brown-Twiss effect using coherent light by lensless holographic projection

Computer-generated holography enables projection of target patterns onto designated planes, providing deterministic control over the probability density function of the projected light intensity. Here, we introduce an active control scheme for the peak value of the Hanbury Brown--Twiss effect, $g^{(2)}(0)$, utilizing lensless holographic projection with coherent light. Notably, single-frame holographic projection yields a markedly different $g^{(2)}(0)$ from its multiframe-averaged counterpart due to the presence of coherent speckle noise. With the coherent speckle noise suppression, we derive an analytical expression $g^{(2)}(0)$ on holographic projection plane, revealing that it is determined by the target coherence length, its statistics, and the numerical aperture of projection system. Our experimental results show good agreement with the theoretical analysis, confirming the joint influence of these factors. By employing dynamic sparse target patterns, we achieve a maximum $g^{(2)}(0)$ of $39.77$. Numerical simulations, benchmarked against experimental measurements, reveal that coherent speckle noise enhances $g^{(2)}(0)$ through mutual superposition with the target pattern, leading to a joint modulation of intensity fluctuations. In summary, by manipulating multiple controllable parameters, we establish a robust strategy for tailoring $g^{(2)}(0)$, paving the way for advanced applications in speckle imaging and optical metrology.

physics.optics↗
arXiv2026

Allo{SR}$^2$: Rectifying One-Step Super-Resolution to Stay Real via Allomorphic Generative Flows

Real-world image super-resolution (Real-SR) has been revolutionized by leveraging the powerful generative priors from Diffusion Models (DMs) and Flow Matching (FM). However, existing one-step methods typically replace Gaussian noise with degraded low-resolution (LR) latents at initialization, introducing a substantial distribution shift that further leads to trajectory deviation and prior collapse under extreme acceleration. To overcome these limitations, we propose Allo{SR}$^2$, a novel FM-based framework that rectifies one-step SR flows via allomorphic generative flows to maintain high-fidelity generative realism. Specifically, we utilize SNR-Guided Trajectory Initialization to identify a statistically aligned intermediate state along the pre-trained path to integrate LR representations into the generative flow. To ensure a stable, low-curvature path for one-step inference, we propose Flow-Anchored Trajectory Consistency (FATC), which explicitly regularizes the velocity field of the underlying probability flow. Furthermore, we develop Allomorphic Trajectory Matching (ATM), a self-adversarial distillation strategy that jointly models the SR flow and the generative flow within a unified velocity field, enabling one-step Real-SR while preserving the generative prior. Extensive experiments on both synthetic and real-world benchmarks demonstrate that Allo{SR}$^2$ achieves state-of-the-art performance in one-step Real-SR, offering a superior balance between fidelity and realism while maintaining extreme efficiency.

cs.CV↗
arXiv2026

Siamese Neural Network for Label-Efficient Critical Phenomena Prediction in 3D Percolation Models

Predicting critical phenomena from limited labeled data remains a challenging task in statistical physics. As percolation theory provides a canonical model for phase transitions with well-established critical exponents, it serves as an ideal benchmark for validating new machine learning frameworks. Here, we introduce a label-efficient learning framework based on a Siamese Neural Network (SNN) to identify phase transitions in three-dimensional site and bond percolation models. Using only 22 labeled probability points drawn entirely from non-critical regions, the method locates percolation thresholds with percent-level accuracy and yields estimates of the critical exponent $ν$ consistent with literature values within statistical uncertainty. Analysis of the learned representations clarifies what the network learns: although trained solely on binary similarity labels, the network autonomously converges to a statistic that coincides quantitatively with the normalized largest-cluster size $S_{max}/L^3$ ($r > 0.99$), the finite-size order parameter of percolation. This underlies the framework's most distinctive capability -- a model trained solely on simple cubic lattices identifies the phase transition in face-centered cubic lattices without retraining. The framework thus offers a complementary route to criticality detection in settings where no quantitative order parameter is explicitly defined or labeled data is scarce.

cond-mat.dis-nn↗
arXiv2026

Dynamic-in-Few-Step: Unifying Dynamic Computation and Few-Step Distillation for Efficient Video Generation

Video Diffusion Models (VDMs) have demonstrated superior generation quality but suffer from prohibitive computational costs. While recent few-step distillation techniques significantly accelerate inference, they typically enforce a static model architecture across all denoising stages, ignoring the varying computational demands inherent to different noise levels. In this work, we propose a novel post-training acceleration framework that exploits this redundancy by integrating dynamic structural sparsification directly into the distillation process. Unlike conventional post-hoc compression applied to a fixed diffusion pipeline, our approach jointly optimizes the denoising steps and structured model sparsity, transforming a pre-trained VDM into a compact, step-specific Mixture-of-Models (MoM). To address the training instability arising from this joint optimization, we introduce a Progressive Training Strategy coupled with an Output Rollout Mechanism, which ensures the coherent learning of structural decisions across timesteps. Furthermore, we develop a specialized inference engine to deploy the resulting MoM efficiently. Our method is orthogonal to existing acceleration techniques and highly effective: On Wan-14B, it removes 24% of the per-step FLOPs on top of 4-step distillation, adding a 1.2x wall-clock gain and reaching a 30x speedup over the 50-step teacher while preserving competitive generation quality.

cs.CV↗
arXiv2026

A Spatial-Spectral-Frequency Interactive Network for Multimodal Remote Sensing Classification

Deep learning-based methods have achieved significant success in remote sensing Earth observation data analysis. Numerous feature fusion techniques address multimodal remote sensing image classification by integrating global and local features. However, these techniques often struggle to extract structural and detail features from heterogeneous and redundant multimodal images. With the goal of introducing frequency domain learning to model key and sparse detail features, this paper introduces the spatial-spectral-frequency interaction network (S$^2$Fin), which integrates pairwise fusion modules across the spatial, spectral, and frequency domains. Specifically, we propose a high-frequency sparse enhancement transformer that employs sparse spatial-spectral attention to optimize the parameters of the high-frequency filter. Subsequently, a two-level spatial-frequency fusion strategy is introduced, comprising an adaptive frequency channel module that fuses low-frequency structures with enhanced high-frequency details, and a high-frequency resonance mask that emphasizes sharp edges via phase similarity. In addition, a spatial-spectral attention fusion module further enhances feature extraction at intermediate layers of the network. Experiments on four benchmark multimodal datasets with limited labeled data demonstrate that S$^2$Fin performs superior classification, outperforming state-of-the-art methods. The code is available at https://github.com/HaoLiu-XDU/SSFin.

cs.CV↗
arXiv2026

Phase diagram of a one-dimensional Ising model with an anomalous Z_2 symmetry

Anomalous global symmetries, which can be realized on the boundary of symmetry-protected topological phases, bring\wlx{s} new phases and phase transitions to condensed matter physics. In this work, we study a one-dimensional model with an anomalous $\mathbb Z_2$ symmetry, using the density-matrix renormalization group method. Besides a symmetry-breaking ferromagnetic phase, we find a gapless phase described by the Tomonaga-Luttinger Liquid theory with the central charge $c=1$. Our numerical finding is compatible of theoretical constraints on possible phases resulting from the symmetry anomaly.

cond-mat.str-el↗
arXiv2026

Beam monitoring for radiotherapy from conventional to FLASH dose rates using Low Gain Avalanche Silicon detectors

We report the performance of low gain avalanche Silicon detectors (LGADs) for instantaneous electron and proton beam monitoring across dose rates ranging from conventional radiotherapy to the FLASH regime, benefiting from the fast response of these detectors of a few nanoseconds. The beam sources provide a dose rate greater than 40~Gy/s through pulses of widths 0.5, 1, 2 and 3~$μ$s for electron beams and 3, 5, 10 $μ$s for proton beams. Two different LGAD devices and silicon diodes are tested, yielding a linear dose response for electron beams up to $\sim$450~Gy/s and for proton beams up to $\sim$12~Gy/s. Beyond the linear regime the response continues to increase with a reduced slope and no true signal plateau is observed, at least up to 1800 Gy/s for electrons and 150 Gy/s for protons. This study contributes towards the instantaneous monitoring of increasingly intense flash beams for radiotherapy using fast detectors such as LGADs since measurements can be performed every fraction of $μ$s.

physics.med-ph↗
arXiv2026

Proximal Mediation Analysis with Unmeasured Treatment-Induced Confounding

Mediation analysis provides a central framework for elucidating causal mechanisms, yet its application is often impeded by treatment-induced confounding, under which the widely used natural mediation effects are generally unidentifiable. Interventional effects have been proposed as an alternative when these confounders are observable; however, identifying and estimating interventional effects remains challenging when confounders are unmeasured. In this paper, we address this issue by using observed variables as proxies for unmeasured treatment-induced confounders. We establish four proximal identification results and develop a multiply robust, semiparametric locally efficient estimator that accommodates flexible machine learning methods for nuisance parameter estimation. The proposed approach is illustrated through simulation studies and a real-data application evaluating racial disparities in life satisfaction mediated by discrimination.

stat.ME↗
arXiv2026

Surface code logical operations on a superconducting quantum processor

Fault-tolerant quantum computation requires logical operations that manipulate encoded information while preserving quantum error-correction protection. In planar surface-code architectures, code deformation and lattice surgery provide a local, measurement-based route to such operations. Here we experimentally realize key elements of patch-based surface-code logical processing on a 107-qubit superconducting quantum processor. We first implement a reusable primitive layer comprising merge and split, patch expansion and shrinkage, and deformations mediated by domain walls and twist defects. We then compose these primitives to realize logical state routing, the logical controlled-NOT gate, and the single-qubit Hadamard and phase gates, which together form a Clifford-generating set. All operations are implemented on distance-three rotated surface-code patches with multi-round syndrome extraction and neural-network decoding, without post-selection. Our results advance superconducting surface-code experiments from protected logical memory to active, patch-based fault-tolerant logical operations.

quant-ph↗
arXiv2026

A Turán Theorem for Cayley Graphs

In this note, we give a Turán theorem for Cayley graphs $\Cay(\Z_p,S)$ over prime cyclic groups $\Z_p$. For a graph $F$ and a finite abelian group $G$, define the Cayley--Turán number by \[ \exCay(F,G) = \max\{|S|:S=-S\subseteq G\setminus\{0\},\ \Cay(G,S)\text{ is }F\text{-free}\}. \] Using a polynomial method, we prove that for every odd prime $p$ and every $1\le r\le p-1$, \[ \exCay(K_{r+1},\Z_p) = p-1-2\left\lfloor\frac{p}{r+1}\right\rfloor . \] The extremal construction is the complement of the short-difference interval \[ D_0=\{0,\pm1,\ldots,\pm\lfloor p/(r+1)\rfloor\}. \] We also discuss what changes for general finite abelian groups, showing why the exact prime-cyclic formula does not extend verbatim to composite cyclic groups.

math.CO↗
arXiv2026

CSD: Content-aware Speculative Decoding for Efficient Image Generation

Speculative decoding (SD) has emerged as a key solution to accelerate the inference of autoregressive models. However, in the field of image generation, it faces the challenge of low acceptance rates, and directly relaxing its criteria leads to degradation in image quality. In this paper, we propose a novel content-aware speculative decoding algorithm, termed CSD, which integrates an entropy-based probability relaxation mechanism with an optimal resampling strategy to enhance the inference efficiency for autoregressive image generation. By leveraging the informational uncertainty inherent in different regions of an image, CSD dynamically adjusts the acceptance probability of candidate tokens, increasing the acceptance rate in low-detail areas to accelerate generation. Moreover, a distribution alignment filter is introduced to ensure the output distribution to be aligned with the target model, which significantly improves the generative quality. Experiments conducted on Lumina-mGPT and Janus-Pro demonstrate that the superiority of the proposed CSD. Our source code is available at https://github.com/aderfebr/CSD.

cs.CV↗