Searcharxiv⌕ Search

arXiv subjects

Xiaolong Liu

Publications and source records attributed to Xiaolong Liu.

At least 19 recordsLinked to original sources
arXiv2026

On Logarithmic Donaldson-Thomas invariants for local Calabi-Yau $4$-folds

In this paper, we study logarithmic Donaldson-Thomas invariants for local log Calabi-Yau $4$-folds with simple normal crossing divisors. Using the family version of shifted Lagrangian classes announced by Khan-Kinjo-Park-Safronov, we construct relative and family logarithmic $\mathsf{DT}_4$-theories for logarithmic Hilbert schemes of curves, and prove a degeneration formula. For logarithmic Hilbert schemes of points, we construct virtual classes and prove a degeneration formula in Chow groups; in particular, it is independent of shifted Lagrangian classes. Finally, combining our degeneration formula with logarithmic cobordism, we explicitly compute the zero-dimensional logarithmic $\mathsf{DT}_4$-invariants for local surface snc pairs.

math.AG↗
arXiv2026

Motus2: A Self-Evolving General World Model for Dexterous Manipulation

General embodied agents should perceive, predict, act, evaluate, and improve within a unified system. World models have shown great promise in building such agents, yet existing models typically append an action output head to a world simulator, without coupling them into a closed decision-and-learning loop for policy improvement. We present Motus2, a self-evolving general world model for dexterous manipulation. Motus2 advances world modeling through model scaling and data scaling. For model scaling, a single model with shared weights exposes three control interfaces: a policy (world-action model), a simulator (action-conditioned world model), and an evaluator (value model). The policy proposes candidate action chunks, the simulator predicts their visual consequences, and the evaluator assesses the predicted outcomes. Their coupling forms a closed decision-and-learning loop for policy improvement. This formulation uses curated expert demonstrations for action learning, while failed and suboptimal interactions provide valuable evidence for dynamics modeling and value learning. For data scaling, Motus2 progresses from large-scale monocular egocentric data to synchronized stereo egocentric data, followed by robot-domain adaptation with robot trajectories and supplementary human-robot alignment data. Motus2 further studies global-autoregressive and hybrid-memory extensions of its sliding-window context, adds tactile feedback for contact-aware control, and is instantiated on a fully biomimetic platform with stereo vision, dual arms, dual dexterous hands, and tactile sensing. Together, egocentric data scaling and closed-loop general world model scaling provide a general path toward self-evolving dexterous manipulation.

cs.RO↗
arXiv2026

Donaldson-Thomas invariants of $[\mathbb C^4/\mathbb Z_r]$

We compute the zero-dimensional Donaldson-Thomas invariants of the quotient stack $[\mathbb{C}^4/\mathbb{Z}_r]$, confirming a conjecture of Cao-Kool-Monavari. Our main theorem is established through an orbifold analogue of Cao-Zhao-Zhou's degeneration formula combined with the zero-dimensional Donaldson-Thomas invariants for $\widetilde{\mathbb C^2/\mathbb Z_r}\times\mathbb{C}^2$ and an explicit determination of orientations of Hilbert schemes of points on $[\mathbb{C}^4/\mathbb{Z}_r]$.

math.AG↗
arXiv2026

ACPO: Asymmetric Credit Policy Optimization via Mode-Local Entropy Surrogate

Outcome-supervised reinforcement learning scales to verifiable reasoning tasks, but trajectory-level rewards assign the same outcome signal to all sampled tokens, overlooking their unequal contributions to the reasoning process. Entropy provides a natural indicator of the model's decision state, yet using it for token-level credit assignment presents two key challenges: long-tail probabilities in large vocabularies corrupt both entropy values and gradients, and uncertainty carries distinct semantics across positive- and non-positive-advantage trajectories. We propose Asymmetric Credit Policy Optimization (ACPO), which replaces global entropy with the complement of the top-token probability as a mode-local proxy. Guided by gradient analysis, ACPO incorporates mismatch routing and saturation correction to shape policy updates into the desired asymmetric form, emphasizing uncertain decisions on positive trajectories while penalizing confident regions on failed ones. Theoretically, ACPO locally preserves the advantage direction while bounding surrogate error. Experiments on mathematical and coding reasoning benchmarks, including AIME 2025 and HumanEval Pro, show that ACPO consistently outperforms both entropy-aware methods (e.g., 80/20, GTPO) and strong outcome-supervised RL baselines (e.g., DAPO, SAPO).

cs.LG↗
arXiv2026

Teaching Tiny VLA Models Where to Look and How to Move

Tiny Vision-Language-Action models are appealing for real-time robotic control, but reducing model scale often weakens two capabilities essential for manipulation: task-conditioned spatial grounding and coherent action generation. We introduce XS-VLA, a lightweight framework that teaches tiny VLA policies "where to look" and "how to move" without increasing deployment-time model cost. For spatial grounding, Coarse-Grained Spatial Distillation uses Qwen3-VL-4B to produce teacher-derived coarse image-plane location labels for task-relevant objects, which are quantized into a spatial vocabulary and distilled into a SmolVLM2-0.25B backbone without human annotations. For action generation, Latent Flow Matching combines a CVAE-style latent variable with flow-based policy learning to organize multimodal demonstrations during training and produce stable action chunks at deployment. On LIBERO, XS-VLA improves average success from 82.8\% to 90.3\% over SmolVLA-0.25B and improves LIBERO-Long from 63.0\% to 89.0\%. In real-world experiments, we design three tasks covering single-arm placement, precision bimanual stacking, and long-horizon sequential coordination, where XS-VLA improves average task success from 21.7\% to 65.0\%. These results show that explicit spatial grounding and latent action-structure learning can make tiny VLA models effective for robotic manipulation.

cs.RO↗
arXiv2026

Dual Inversion for Text-to-Image Diffusion Models: From Both Prompt and Noise Perspectives

Prompt inversion, as a typical reverse engineering technique, enables text-to-image (T2I) diffusion models to generate the desired target images without extensive prompt engineering. However, existing prompt inversion methods suffer from significant limitations: (1) gradient-based methods are unstable and uninterpretable, often resulting in generated images with severe artifacts; (2) gradient-free methods yield human-readable prompts but still fail to preserve visual fidelity due to the lack of fine-grained detail alignment. We contend that the limitations stem from treating prompt inversion as a sufficient condition for reverse engineering, ignoring the critical role of the latent noise that encodes structural information. Consequently, we propose Dualin (Dual inversion), a two-stage method that jointly recovers both the semantic prompt and latent noise of the target image. In the first stage, we integrate vision-language model, CLIP and large language model to invert a faithful, human-interpretable hard prompt. In the second stage, unconditional DDIM inversion reconstructs the exact latent noise of the target image, guaranteeing the consistency at the structural information level. Theoretically, we prove that the inverted noise enables flexible image editing without re-optimization. Extensive experiments on diverse datasets demonstrate that Dualin simultaneously generates high-quality inverted prompts and achieves state-of-the-art image fidelity. Additionally, Dualin can establish a robust foundation for the precise and controllable image editing.

cs.CV↗
arXiv2026

Thinking in Video: Can Video Generators Really Reason About the Real World?

Recent advances in world models and video generation have given rise to an emerging reasoning paradigm that leverages video generative models to simulate, predict, and reason about real-world dynamics. We redefine this paradigm as Thinking in Video, where video is not merely an output artifact but a medium for constructing, extending, and verifying causal thought. However, this promise remains unverified: convincing rollouts may reflect memorized appearances rather than causal understanding, while existing metrics separate perceptual fidelity from semantic logic. To evaluate whether video generators support such reasoning, we introduce the Causal-Generative Dual-Judge (CGDJ), auditing World Model Consistency from two perspectives. Explicit Causal Perception tests whether a generator reads a video scenario as a reasoning problem through spatio-temporal flattened visual question answering, while Implicit Generative Perception-Prediction Gap evaluates whether it renders the causal consequence as a consistent future video. Applying CGDJ to representative open- and closed-source generators reveals a clear Perception-Prediction Gap: open-source models produce plausible dynamics despite near-zero explicit causal perception, whereas advanced closed-source systems show stronger but still limited alignment between reasoning and generation. Further analysis exposes audio-visual misalignment, where models verbalize correct causal logic more reliably than they render it, challenging the "world simulator" narrative.

cs.CV↗
arXiv2026

Performance Characterization of Frequency-Selective Wireless Power Transfer Toward Scalable Untethered Magnetic Actuation

Frequency-selective wireless power transfer provides a feasible route to enable independent actuation and control of multiple untethered robots in a common workspace; however, the scalability remains unquantified, particularly the maximum number of resonators that can be reliably addressed within a given frequency bandwidth. To address this, we formulate the relationship between resonator quality factor (Q-factor) and the number of individually addressable inductor-capacitor (LC) resonant energy harvesters within a fixed radio-frequency (RF) spectrum, and we convert selectively activated harvested energy into mechanical motion. We theoretically proved and experimentally demonstrated that scalability depends primarily on the Q-factor. For this proof-of-concept study, we define effective series resistance as a function of frequency allocating bandwidths to discrete actuators. We provide design equations for scaling untethered magnetic actuation with Q-factor optimization. Resonator networks spanning bandwidths from 100kHz to 1MHz were analyzed to quantify how increasing the number of resonators affects independent addressability. We validated the approach experimentally by fabricating three centimeter-scale untethered actuators that selectively trigger the motion of mechanical beams at 734kHz, 785kHz, and 855kHz. We also characterized the generated mechanical force and the activation bandwidth of each actuator, confirming that no unintended cross-triggering occurred.

eess.SY↗
arXiv2026

ACE: Agentic Control for Embodied Manipulation via Zero-shot Workflow Reasoning

Open-ended tabletop manipulation requires agents to not only understand natural language but also adapt to dynamic environments and execution failures. We present ACE (Agentic Control for Embodied Manipulation), a zero-shot workflow reasoning framework for tabletop pick-and-place from natural language. Rather than relying on direct low-level action mapping, ACE combines agentic workflow reasoning with two robot-facing executable skills: a visual grounding interface and a reusable pick-and-place primitive. To bridge semantic reasoning and physical control, the active sub-goal is grounded into a mask-mediated vision-action interface. This unified mask specifies the target object and destination, is tracked over time, exposed for human verification, and ultimately passed to a task-agnostic downstream policy for execution. Crucially, ACE operates in a closed loop supported by a multi-timescale memory. After an action is executed, the system automatically verifies whether the intended sub-goal succeeded, using the outcome to advance, retry, repair, or replan. This enables online adaptation to user corrections, scene changes, and physical failures. We evaluate ACE on logically complex, long-horizon tasks, including zero-shot multi-step equation formation with number cubes and constraint-based object retrieval. ACE demonstrates task-level zero-shot generalization on novel semantic constraints and randomized tabletop scenes without task-specific retraining. Specifically, while standard end-to-end baselines struggle to complete these logically demanding tasks, ACE achieves a 50% success rate in equation formation and a 70% success rate in constraint retrieval. This contrast demonstrates that explicit workflow reasoning and mask-mediated control offer a robust, practical route toward adaptable robotic manipulation.

cs.RO↗
arXiv2026

Kunlun: Establishing Scaling Laws for Massive-Scale Recommendation Systems through Unified Architecture Design

Deriving predictable scaling laws that govern the relationship between model performance and computational investment is crucial for designing and allocating resources in massive-scale recommendation systems. While such laws are established for large language models, they remain challenging for recommendation systems, especially those processing both user history and context features. We identify poor scaling efficiency as the main barrier to predictable power-law scaling, stemming from inefficient modules with low Model FLOPs Utilization (MFU) and suboptimal resource allocation. We introduce Kunlun, a scalable architecture that systematically improves model efficiency and resource allocation. Our low-level optimizations include Generalized Dot-Product Attention (GDPA), Hierarchical Seed Pooling (HSP), and Sliding Window Attention. Our high-level innovations feature Computation Skip (CompSkip) and Event-level Personalization. These advances increase MFU from 17% to 37% on NVIDIA B200 GPUs and double scaling efficiency over state-of-the-art methods. Kunlun is now deployed in major Meta Ads models, delivering significant production impact.

cs.IR↗
arXiv2026

LoopFM: Learning frOm HistOrical RePresentations of Foundation Model for Recommendation

Knowledge distillation (KD) transfers a single scalar prediction from a large foundation model (FM) to compact vertical models (VMs), suffering from diminishing transfer ratio -- the fraction of FM improvement captured by the VM -- as a single scalar cannot convey the rich intermediate knowledge that larger FMs learn. To address this bottleneck, we propose LoopFM (Learning frOm HistOrical RePresentations of FM), a framework that opens a high-bandwidth transfer channel by structuring FM intermediate embeddings as input features (e.g., user history sequence) for downstream VMs, without requiring real-time FM inference at serving and architectural coupling between FM and VM. We provide a theoretical framework for LoopFM with a gain decomposition and transfer-ratio analysis. On three public benchmarks, LoopFM demonstrates strong AUC improvements (e.g., 6%+ on TaobaoAd) and complementary knowledge transfer capability with KD. On industrial-scale systems (billions of examples, trillion-parameter FMs), LoopFM approximately doubles the knowledge transfer ratio on top of KD, delivering a +0.5% conversion improvement in the first half after its initial launch, and +1.03% and +1.22% conversion improvement from two individual launches in the subsequent half.

cs.LG↗
arXiv2026

Hyperdoped silicon photodetectors enable room-temperature computational SWIR imaging at 1550 nm

Silicon's bandgap inherently restricts its photodetection to wavelengths below 1100 nm, necessitating the integration of costly III-V semiconductors for short-wave infrared applications. Hyperdoping silicon beyond the solid solubility limit offers a promising "silicon-native" alternative, yet achieving practical short-wave infrared applications at room temperature remains a formidable challenge. Here, we demonstrate a high-detectivity hyperdoped silicon photodetector enabling room-temperature computational short-wave infrared imaging beyond Si bandgap wavelength at λ = 1550 nm. By integrating an ultrafast laser heating process step to reduce the dark current while keeping high responsivity, we achieve a specific detectivity D^* exceeding 10^9 Jones for 1550 nm at room temperature working in a forward-biased, photoconductive mode. The improved detectivity, coupled with a 59.4 dB linear dynamic range and kHz-scale bandwidth, allows us to demonstrate a single-pixel imaging system that reconstructs 1550 nm scenes at 65x63 pixels without cryogenic cooling. Our devices simultaneously support visible-light imaging, offering a path toward monolithically integrated, multispectral Si-native optical sensors. These results establish ultrafast-laser hyperdoped silicon as a viable platform for low-cost, room-temperature, short-wave infrared photonics, bridging the gap between advanced materials science and practical computational imaging system.

physics.optics↗
arXiv2026

The Perceptual Bandwidth Bottleneck in Vision-Language Models: Active Visual Reasoning via Sequential Experimental Design

Visual perception in modern Vision-Language Models (VLMs) is constrained by a perceptual bandwidth bottleneck: a broad field of view preserves global context but sacrifices the fine-grained details required for complex reasoning. We argue that high-resolution visual reasoning is therefore not only semantic reasoning but also task-relevant evidence acquisition under limited perceptual bandwidth. Inspired by active vision and information foraging, we formalise this process as sequential Bayesian optimal experimental design (S-BOED), where an agent decides which visual evidence to acquire before answering. Since exact Bayesian inference is intractable in continuous gigapixel spaces, we derive a tractable coverage--resolution objective as a proxy for task-relevant information gain. We instantiate this framework with FOVEA, a training-free procedure that refines VLM crop proposals through evidence-oriented probing. Experiments on high-resolution benchmarks show consistent gains over direct and ReAct-style baselines, with particularly strong improvements in search-dominated remote-sensing settings.

cs.CV↗
arXiv2026

Chain-of-Models Pre-Training: Rethinking Training Acceleration of Vision Foundation Models

In this paper, we present Chain-of-Models Pre-Training (CoM-PT), a novel performance-lossless training acceleration method for vision foundation models (VFMs). This approach fundamentally differs from existing acceleration methods in its core motivation: rather than optimizing each model individually, CoM-PT is designed to accelerate the training pipeline at the model family level, scaling efficiently as the model family expands. Specifically, CoM-PT establishes a pre-training sequence for the model family, arranged in ascending order of model size, called model chain. In this chain, only the smallest model undergoes standard individual pre-training, while the other models are efficiently trained through sequential inverse knowledge transfer from their smaller predecessors by jointly reusing the knowledge in the parameter space and the feature space. As a result, CoM-PT enables all models to achieve performance that is mostly superior to standard individual training while significantly reducing training cost, and this is extensively validated across 45 datasets spanning zero-shot and fine-tuning tasks. Notably, its efficient scaling property yields a remarkable phenomenon: training more models even results in higher efficiency. For instance, when pre-training on CC3M: i) given ViT-L as the largest model, progressively prepending smaller models to the model chain reduces computational complexity by up to 72%; ii) within a fixed model size range, as the VFM family scales across 3, 4, and 7 models, the acceleration ratio of CoM-PT exhibits a striking leap: from 4.13X to 5.68X and 7.09X. Since CoM-PT is naturally agnostic to specific pre-training paradigms, we open-source the code to spur further extensions in more computationally intensive scenarios, such as large language model pre-training.

cs.CV↗
arXiv2026

Nanostructuring SiC by sequential plasma oxidation and reactive ion etching

Silicon carbide (SiC) is a highly promising material for the rapidly growing UV detection industry due to its visible-blindness, low dark current, and exceptional thermal and chemical stability. Despite these advantages, the performance of state-of-the-art SiC UV detectors remains limited due to high reflectance losses, even with the use of anti-reflection coatings. Here, we develop a reactive ion etching process for nanostructuring SiC to eliminate the reflectance losses. The process is based on consecutive oxidation and etching cycles. Consequently, a reflectance below 0.5% is achieved from deep UV (200 nm) to close to the SiC cut-off (~360 nm). The nanostructures are effective even at large incident angles as the reflectance remains practically unchanged up to 60 degrees. Furthermore, it is confirmed that the process consumes only ~1 um of SiC and is compatible with Al2O3 masking, thereby facilitating straightforward integration into device fabrication. The developed cyclical etching process could also prove useful for SiC etching in general.

physics.app-ph↗
arXiv2026

Directional Manipulation of a Staggered Charge Density Wave and Kondo Resonance in UTe2

UTe2 is a rare example of a correlated quantum material in which unconventional density wave orders, Kondo physics, spin-triplet pairing, and reentrant superconductivity coexist within the same electronic system. Its superconducting state develops out of a strongly correlated normal phase. The identification and control of competing or intertwined normal-state orders are thus central to elucidating the electronic landscape from which its superconductivity arises. Here, using scanning tunnelling microscopy (STM) in a vector magnetic field, we uncover a previously unreported staggered charge-density-wave (CDW) in high-quality UTe2 crystals and demonstrate its strong directional response to an external magnetic field: the staggered CDW is completely quenched by a modest 1.7 T field aligned with the quasi-one-dimensional uranium chain direction (a-axis), while remaining robust against fields along other crystallographic directions. This pronounced anisotropy is consistent with an orbital-driven mechanism that leads to a field-tuned quantum phase transition. Strikingly and counterintuitively, the same field orientation and strength concomitantly alter the hybridization gap and suppress the 5f Kondo resonance. Modelling indicates that this correlated evolution arises from a switch of the dominant hybridization channel from Te 5p- U 5f to U 6d- U 5f coupling, suggesting an intimate interplay between CDW and the Kondo effect. Our work establishes an effective tuning knob for the intertwined orders in UTe2 and provides evidence for orbital-selective Kondo hybridization, shedding light on its correlated normal state.

cond-mat.str-el↗
arXiv2026

Full-channel wavefront manipulation of surface waves with chirality-assisted geometric-phase metasurface

Owing to their localized field enhancement and subwavelength resolution, surface waves (SWs) offer broad application potential in communications, sensing, and photonics via on-chip wavefront manipulation. This makes multi-channel SW wavefront manipulation highly desirable. However, conventional metasurfaces for SW wavefront shaping, relying on geometric and propagation phase mechanisms, typically exhibit similar functionalities for co- or cross-polarized output channels under different circularly polarized (CP) incidences, thereby limiting the development of high-capacity on-chip integrated devices. Here, by introducing the chirality-assisted phase as an additional phase control mechanism, we effectively decouple both co- and cross-polarized output channels, enabling independent SW wavefront shaping in four distinct channels. We numerically and experimentally demonstrate two metasurfaces in the microwave range: a four-channel SW meta-deflector and a four-channel SW metadevice that simultaneously produces a focused SW beam, a SW Bessel beam, and two deflected SW beams in different directions. Therefore, chirality-assisted geometric-phase metasurfaces provide a versatile platform for multi-channel SW wavefront engineering, offering significant potential for high-capacity on-chip communication and integrated photonic systems.

physics.optics↗
arXiv2026

Quasiparticle Interference Kernel Extraction with Variational Autoencoders via Latent Alignment

Quasiparticle interference (QPI) imaging is a powerful tool for probing electronic structures in quantum materials, but extracting the single-scatterer QPI pattern (i.e., the kernel) from a multi-scatterer image remains a fundamentally ill-posed inverse problem, because many different kernels can combine to produce almost the same observed image, and noise or overlaps further obscure the true signal. Existing solutions to this extraction problem rely on manually zooming into small local regions with isolated single-scatterers. This is infeasible for real cases where scattering conditions are too complex. In this work, we propose the first AI-based framework for QPI kernel extraction, which models the space of physically valid kernels and uses this knowledge to guide the inverse mapping. We introduce a two-step learning strategy that decouples kernel representation learning from observation-to-kernel inference. In the first step, we train a variational autoencoder to learn a compact latent space of scattering kernels. In the second step, we align the latent representation of QPI observations with those of the pre-learned kernels using a dedicated encoder. This design enables the model to infer kernels robustly under complex, entangled scattering conditions. We construct a diverse and physically realistic QPI dataset comprising 100 unique kernels and evaluate our method against a direct one-step baseline. Experimental results demonstrate that our approach achieves significantly higher extraction accuracy, improved generalization to unseen kernels. To further validate its effectiveness, we also apply the method to real QPI data from Ag and FeSe samples, where it reliably extracts meaningful kernels under complex scattering conditions.

cs.LG↗