Searcharxiv⌕ Search

arXiv subjects

Lei Lei

Publications and source records attributed to Lei Lei.

At least 37 records · Page 2Linked to original sources
arXiv2026

OpAgent: Operator Agent for Web Navigation

To fulfill user instructions, autonomous web agents must contend with the inherent complexity and volatile nature of real-world websites. Conventional paradigms predominantly rely on Supervised Fine-Tuning (SFT) or Offline Reinforcement Learning (RL) using static datasets. However, these methods suffer from severe distributional shifts, as offline trajectories fail to capture the stochastic state transitions and real-time feedback of unconstrained wide web environments. In this paper, we propose a robust Online Reinforcement Learning WebAgent, designed to optimize its policy through direct, iterative interactions with unconstrained wide websites. Our approach comprises three core innovations: 1) Hierarchical Multi-Task Fine-tuning: We curate a comprehensive mixture of datasets categorized by functional primitives -- Planning, Acting, and Grounding -- establishing a Vision-Language Model (VLM) with strong instruction-following capabilities for Web GUI tasks. 2) Online Agentic RL in the Wild: We develop an online interaction environment and fine-tune the VLM using a specialized RL pipeline. We introduce a Hybrid Reward Mechanism that combines a ground-truth-agnostic WebJudge for holistic outcome assessment with a Rule-based Decision Tree (RDT) for progress reward. This system effectively mitigates the credit assignment challenge in long-horizon navigation. Notably, our RL-enhanced model achieves a 38.1\% success rate (pass@5) on WebArena, outperforming all existing monolithic baselines. 3) Operator Agent: We introduce a modular agentic framework, namely \textbf{OpAgent}, orchestrating a Planner, Grounder, Reflector, and Summarizer. This synergy enables robust error recovery and self-correction, elevating the agent's performance to a new State-of-the-Art (SOTA) success rate of \textbf{71.6\%}.

cs.AI↗
arXiv2026

MiroThinker: Pushing the Performance Boundaries of Open-Source Research Agents via Model, Context, and Interactive Scaling

We present MiroThinker v1.0, an open-source research agent designed to advance tool-augmented reasoning and information-seeking capabilities. Unlike previous agents that only scale up model size or context length, MiroThinker explores interaction scaling at the model level, systematically training the model to handle deeper and more frequent agent-environment interactions as a third dimension of performance improvement. Unlike LLM test-time scaling, which operates in isolation and risks degradation with longer reasoning chains, interactive scaling leverages environment feedback and external information acquisition to correct errors and refine trajectories. Through reinforcement learning, the model achieves efficient interaction scaling: with a 256K context window, it can perform up to 600 tool calls per task, enabling sustained multi-turn reasoning and complex real-world research workflows. Across four representative benchmarks-GAIA, HLE, BrowseComp, and BrowseComp-ZH-the 72B variant achieves up to 81.9%, 37.7%, 47.1%, and 55.6% accuracy respectively, surpassing previous open-source agents and approaching commercial counterparts such as GPT-5-high. Our analysis reveals that MiroThinker benefits from interactive scaling consistently: research performance improves predictably as the model engages in deeper and more frequent agent-environment interactions, demonstrating that interaction depth exhibits scaling behaviors analogous to model size and context length. These findings establish interaction scaling as a third critical dimension for building next-generation open research agents, complementing model capacity and context windows.

cs.CL↗
arXiv2026

New $H(z)$ measurement at Redshift = 0.12 with DESI Data Release 1

The Hubble parameter ($H(z)$) is a function of the redshift and a reliable measurement is very important to understand the expansion history of the Universe. In this work, we perform full-spectrum fitting using BAGPIPES on more than four thousand massive, passively evolving galaxies released by the DESI collaboration to estimate their cosmological-independent stellar ages and star-formation histories, and derive a new measurement of $H(z=0.12)=71.33 \pm 4.20~{\rm km~s^{-1}~Mpc^{-1}}$, which is well consistent with those derived in other ways.

astro-ph.CO↗
arXiv2026

NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models: Datasets, Methods and Results

This paper presents an overview of the NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models. This challenge utilizes a new short-form UGC (S-UGC) video restoration benchmark, termed KwaiVIR, which is contributed by USTC and Kuaishou Technology. It contains both synthetically distorted videos and real-world short-form UGC videos in the wild. For this edition, the released data include 200 synthetic training videos, 48 wild training videos, 11 validation videos, and 20 testing videos. The primary goal of this challenge is to establish a strong and practical benchmark for restoring short-form UGC videos under complex real-world degradations, especially in the emerging paradigm of generative-model-based S-UGC video restoration. This challenge has two tracks: (i) the primary track is a subjective track, where the evaluation is based on a user study; (ii) the second track is an objective track. These two tracks enable a comprehensive assessment of restoration quality. In total, 95 teams have registered for this competition. And 12 teams submitted valid final solutions and fact sheets for the testing phase. The submitted methods achieved strong performance on the KwaiVIR benchmark, demonstrating encouraging progress in short-form UGC video restoration in the wild.

cs.CV↗
arXiv2026

Instructing LLMs to Negotiate using Reinforcement Learning with Verifiable Rewards

The recent advancement of Large Language Models (LLMs) has established their potential as autonomous interactive agents. However, they often struggle in strategic games of incomplete information, such as bilateral price negotiation. In this paper, we investigate if Reinforcement Learning from Verifiable Rewards (RLVR) can effectively teach LLMs to negotiate. Specifically, we explore the strategic behaviors that emerge during the learning process. We introduce a framework that trains a mid-sized buyer agent against a regulated LLM seller across a wide distribution of real-world products. By grounding reward signals directly in the maximization of economic surplus and strict adherence to private budget constraints, we reveal a novel four-phase strategic evolution. The agent progresses from naive bargaining to using aggressive starting prices, moves through a phase of deadlock, and ultimately develops sophisticated persuasive skills. Our results demonstrate that this verifiable training allows a 30B agent to significantly outperform frontier models over ten times its size in extracting surplus. Furthermore, the trained agent generalizes robustly to stronger counterparties unseen during training and remains effective even when facing hostile, adversarial seller personas.

cs.AI↗
arXiv2026

DMAligner: Enhancing Image Alignment via Diffusion Model Based View Synthesis

Image alignment is a fundamental task in computer vision with broad applications. Existing methods predominantly employ optical flow-based image warping. However, this technique is susceptible to common challenges such as occlusions and illumination variations, leading to degraded alignment visual quality and compromised accuracy in downstream tasks. In this paper, we present DMAligner, a diffusion-based framework for image alignment through alignment-oriented view synthesis. DMAligner is crafted to tackle the challenges in image alignment from a new perspective, employing a generation-based solution that showcases strong capabilities and avoids the problems associated with flow-based image warping. Specifically, we propose a Dynamics-aware Diffusion Training approach for learning conditional image generation, synthesizing a novel view for image alignment. This incorporates a Dynamics-aware Mask Producing (DMP) module to adaptively distinguish dynamic foreground regions from static backgrounds, enabling the diffusion model to more effectively handle challenges that classical methods struggle to solve. Furthermore, we develop the Dynamic Scene Image Alignment (DSIA) dataset using Blender, which includes 1,033 indoor and outdoor scenes with over 30K image pairs tailored for image alignment. Extensive experimental results demonstrate the superiority of the proposed approach on DSIA benchmarks, as well as on a series of widely-used video datasets for qualitative comparisons. Our code is available at https://github.com/boomluo02/DMAligner.

cs.CV↗
arXiv2026

Stringent constraint on the CCC+TL cosmology with $H(z)$ Measurements

Recently, the Covarying Coupling Constants and Tired Light (CCC+TL) hybrid model was proposed to explain the unexpectedly small angular diameters of high-redshift galaxies observed by the James Webb Space Telescope (JWST) that are challenging to reconcile with the $Λ$CDM model. In this work, we test the CCC+TL model against model-independent Hubble parameter [$H(z)$] measurements obtained from cosmic chronometers. It turns out that the parameter set optimized for the type-Ia supernova (SN Ia) dataset within the CCC+TL model fails to reproduce the $H(z)$ data, but the $Λ$CDM model works well. Statistical comparison using the $Δχ^2$ strongly favors $Λ$CDM over CCC+TL for the $H(z)$ data, with $Δχ^2 = 61.52$. Crucially, the CCC+TL framework exhibits a severe internal tension, where the SN Ia-optimized speed-of-light variation index $α$ is rejected by the $H(z)$ dataset with a likelihood ratio of $\it{R} \approx 1.7 \times 10^{-14}$. Our result suggests that the tension posed by JWST observations of compact high-$z$ galaxies may originate from the intrinsic properties and evolution of galaxies in the early universe.

astro-ph.CO↗
arXiv2026

MiM-DiT: MoE in MoE with Diffusion Transformers for All-in-One Image Restoration

All-in-one image restoration is challenging because different degradation types, such as haze, blur, noise, and low-light, impose diverse requirements on restoration strategies, making it difficult for a single model to handle them effectively. In this paper, we propose a unified image restoration framework that integrates a dual-level Mixture-of-Experts (MoE) architecture with a pretrained diffusion model. The framework operates at two levels: the Inter-MoE layer adaptively combines expert groups to handle major degradation types, while the Intra-MoE layer further selects specialized sub-experts to address fine-grained variations within each type. This design enables the model to achieve coarse-grained adaptation across diverse degradation categories while performing fine-grained modulation for specific intra-class variations, ensuring both high specialization in handling complex, real-world corruptions. Extensive experiments demonstrate that the proposed method performs favorably against the state-of-the-art approaches on multiple image restoration task.

cs.CV↗
arXiv2026

Possible evidence for extended X-ray emission surrounding PSR B0656+14 with eROSITA

Extended very-high-energy $γ$-ray emission from middle-aged pulsars as revealed recently by several groundbased $γ$-ray experiments has strong implication on the transport of high-energy particles in the interstellar medium surrounding those pulsars. The $γ$-ray emission is widely believed to be produced by high-energy electrons and positrons accelerated by the pulsar wind nebulae when scattering off the interstellar radiation field via the inverse Compton process. Consequently, multiwavelength counterparts of the $γ$-ray halos are expected to be present, which have not yet been detected. In this work we report the possible detection of extended X-ray emission from a $\sim 0.2\degr$ radius region around PSR B0656+14 with eROSITA. In spite that there are uncertainties of the on-orbit point spread function of the pointing mode, the radial profile of PSR B0656+14 is found to be broader than that of a star at similar observational conditions, indicating that emission is possibly from the expected extended halo around the pulsar. The spectrum of the emission can be described by a power-law function with an index of $\sim3.7$. Its surface brightness declines with radius faster than the prediction of the particle diffusion and synchrotron radiation in a uniform magnetic field, suggesting the existence of a radial gradient of the magnetic field strength as $\sim r^{-1}$. The magnetic field strength in the X-ray emitting region is constrained to be $4-10~μ$G.

astro-ph.HE↗
arXiv2026

RAW-Flow: Advancing RGB-to-RAW Image Reconstruction with Deterministic Latent Flow Matching

RGB-to-RAW reconstruction, or the reverse modeling of a camera Image Signal Processing (ISP) pipeline, aims to recover high-fidelity RAW data from RGB images. Despite notable progress, existing learning-based methods typically treat this task as a direct regression objective and struggle with detail inconsistency and color deviation, due to the ill-posed nature of inverse ISP and the inherent information loss in quantized RGB images. To address these limitations, we pioneer a generative perspective by reformulating RGB-to-RAW reconstruction as a deterministic latent transport problem and introduce a novel framework named RAW-Flow, which leverages flow matching to learn a deterministic vector field in latent space, to effectively bridge the gap between RGB and RAW representations and enable accurate reconstruction of structural details and color information. To further enhance latent transport, we introduce a cross-scale context guidance module that injects hierarchical RGB features into the flow estimation process. Moreover, we design a dual-domain latent autoencoder with a feature alignment constraint to support the proposed latent transport framework, which jointly encodes RGB and RAW inputs while promoting stable training and high-fidelity reconstruction. Extensive experiments demonstrate that RAW-Flow outperforms state-of-the-art approaches both quantitatively and visually.

cs.CV↗
arXiv2026

Evidence for GeV emission of the superluminous supernova SN 2017egm

Superluminous supernovae (SLSNe) are a new class of transients with luminosities $\sim10 -100$ times larger than the usual core-collapse supernovae (SNe). Their origin is still unclear and one widely discussed scenario involves a millisecond magnetar central engine. The GeV-TeV emission of SLSNe has been predicted in the literature but has not been convincingly detected yet. Here we report the results of the search for $γ$-ray emission in the direction of SN 2017egm, one of the closest SLSNe detected so far, using 15 years of {\it Fermi}-LAT Pass 8 data. There is a transient $γ$-ray source appearing about 2 months after this event and lasting a few months. Monte Carlo simulations show that the $γ$-ray signal has a global significance of {\it at least} 4$σ$. Both the peak time and the luminosity of the GeV emission are consistent with the magnetar model prediction, suggesting that such a GeV transient is the high-energy counterpart of SN 2017egm and the central engine of this SLSNe is a young magnetar.

astro-ph.HE↗
arXiv2026

Unifying Appearance Codes and Bilateral Grids for Driving Scene Gaussian Splatting

Neural rendering techniques, including NeRF and Gaussian Splatting (GS), rely on photometric consistency to produce high-quality reconstructions. However, in real-world scenarios, it is challenging to guarantee perfect photometric consistency in acquired images. Appearance codes have been widely used to address this issue, but their modeling capability is limited, as a single code is applied to the entire image. Recently, the bilateral grid was introduced to perform pixel-wise color mapping, but it is difficult to optimize and constrain effectively. In this paper, we propose a novel multi-scale bilateral grid that unifies appearance codes and bilateral grids. We demonstrate that this approach significantly improves geometric accuracy in dynamic, decoupled autonomous driving scene reconstruction, outperforming both appearance codes and bilateral grids. This is crucial for autonomous driving, where accurate geometry is important for obstacle avoidance and control. Our method shows strong results across four datasets: Waymo, NuScenes, Argoverse, and PandaSet. We further demonstrate that the improvement in geometry is driven by the multi-scale bilateral grid, which effectively reduces floaters caused by photometric inconsistency.

cs.CV↗
arXiv2026

Physics-embedded neural computational electron microscopy for quantitative 4D nanometrology

The fusion of rigorous physical laws with flexible data-driven learning represents a new frontier in scientific simulation, yet bridging the gap between physical interpretability and computational efficiency remains a grand challenge. In electron microscopy, this divide limits the ability to quantify three-dimensional topography from two-dimensional projections, fundamentally constraining our understanding of nanoscale structure-function relationships. Here, we present a physics-embedded neural computational microscopy framework that achieves metrological three-dimensional reconstruction by deeply coupling a differentiable electron-optical forward model with deep learning. By introducing a Vision Field Transformer as a high-speed, differentiable surrogate for physical process analysis simulations, we establish an end-to-end, self-supervised optimization loop that enforces strict physical consistency with hardware geometry. This synergy enables single-shot, quantitative three-dimensional nanometrology with precision comparable to atomic force microscopy but at orders of magnitude higher throughput. Furthermore, we demonstrate the capability for four-dimensional (3D real space plus time) in situ characterization by tracking the dynamic evolution of surface nanostructure during copper redox, revealing hidden crystallographic kinetics invisible to conventional imaging. Our work not only redefines the limits of scanning electron microscopy but also establishes a generalizable archetype for solving ill-posed inverse problems across physical sciences, unlocking the full potential of simulation as a third pillar of discovery.

physics.optics↗
arXiv2025

Step-GUI Technical Report

Recent advances in multimodal large language models unlock unprecedented opportunities for GUI automation. However, a fundamental challenge remains: how to efficiently acquire high-quality training data while maintaining annotation reliability? We introduce a self-evolving training pipeline powered by the Calibrated Step Reward System, which converts model-generated trajectories into reliable training signals through trajectory-level calibration, achieving >90% annotation accuracy with 10-100x lower cost. Leveraging this pipeline, we introduce Step-GUI, a family of models (4B/8B) that achieves state-of-the-art GUI performance (8B: 80.2% AndroidWorld, 48.5% OSWorld, 62.6% ScreenShot-Pro) while maintaining robust general capabilities. As GUI agent capabilities improve, practical deployment demands standardized interfaces across heterogeneous devices while protecting user privacy. To this end, we propose GUI-MCP, the first Model Context Protocol for GUI automation with hierarchical architecture that combines low-level atomic operations and high-level task delegation to local specialist models, enabling high-privacy execution where sensitive data stays on-device. Finally, to assess whether agents can handle authentic everyday usage, we introduce AndroidDaily, a benchmark grounded in real-world mobile usage patterns with 3146 static actions and 235 end-to-end tasks across high-frequency daily scenarios (8B: static 89.91%, end-to-end 52.50%). Our work advances the development of practical GUI agents and demonstrates strong potential for real-world deployment in everyday digital interactions.

cs.CV↗
arXiv2025

Efficient stereo matching on embedded GPUs with zero-means cross correlation

Mobile stereo-matching systems have become an important part of many applications, such as automated-driving vehicles and autonomous robots. Accurate stereo-matching methods usually lead to high computational complexity; however, mobile platforms have only limited hardware resources to keep their power consumption low; this makes it difficult to maintain both an acceptable processing speed and accuracy on mobile platforms. To resolve this trade-off, we herein propose a novel acceleration approach for the well-known zero-means normalized cross correlation (ZNCC) matching cost calculation algorithm on a Jetson Tx2 embedded GPU. In our method for accelerating ZNCC, target images are scanned in a zigzag fashion to efficiently reuse one pixel's computation for its neighboring pixels; this reduces the amount of data transmission and increases the utilization of on-chip registers, thus increasing the processing speed. As a result, our method is 2X faster than the traditional image scanning method, and 26% faster than the latest NCC method. By combining this technique with the domain transformation (DT) algorithm, our system show real-time processing speed of 32 fps, on a Jetson Tx2 GPU for 1,280x384 pixel images with a maximum disparity of 128. Additionally, the evaluation results on the KITTI 2015 benchmark show that our combined system is more accurate than the same algorithm combined with census by 7.26%, while maintaining almost the same processing speed. Source Code: https://github.com/changqiong/Z2ZNCC.git

cs.CV↗
arXiv2025

Safe and Sustainable Electric Bus Charging Scheduling with Constrained Hierarchical DRL

The integration of Electric Buses (EBs) with renewable energy sources such as photovoltaic (PV) panels is a promising approach to promote sustainable and low-carbon public transportation. However, optimizing EB charging schedules to minimize operational costs while ensuring safe operation without battery depletion remains challenging - especially under real-world conditions, where uncertainties in PV generation, dynamic electricity prices, variable travel times, and limited charging infrastructure must be accounted for. In this paper, we propose a safe Hierarchical Deep Reinforcement Learning (HDRL) framework for solving the EB Charging Scheduling Problem (EBCSP) under multi-source uncertainties. We formulate the problem as a Constrained Markov Decision Process (CMDP) with options to enable temporally abstract decision-making. We develop a novel HDRL algorithm, namely Double Actor-Critic Multi-Agent Proximal Policy Optimization Lagrangian (DAC-MAPPO-Lagrangian), which integrates Lagrangian relaxation into the Double Actor-Critic (DAC) framework. At the high level, we adopt a centralized PPO-Lagrangian algorithm to learn safe charger allocation policies. At the low level, we incorporate MAPPO-Lagrangian to learn decentralized charging power decisions under the Centralized Training and Decentralized Execution (CTDE) paradigm. Extensive experiments with real-world data demonstrate that the proposed approach outperforms existing baselines in both cost minimization and safety compliance, while maintaining fast convergence speed.

cs.LG↗
arXiv2025

Lensing amplitude anomaly and varying electron mass alleviate the Hubble and $S_8$ tensions

Cosmological measurements have revealed tensions within the standard $Λ$CDM model, notably discrepancies in the Hubble constant and $S_8$ parameter. A modified recombination scenario involving a time-varying electron mass has been proposed as a feasible solution to the Hubble tension without exacerbating the $S_8$ tension. Recent observations have further revealed other potential deviations from the $Λ$CDM framework, such as non-flat spatial curvature and an anomalous CMB lensing amplitude. In this study, we explore whether introducing a variation in the electron mass $m_e$, allowing non-zero spatial curvature $Ω_K$, and a free lensing amplitude $A_{\rm lens}$ can resolve these persistent tensions. Using the Planck Public Release (PR) 3 and ACT power spectra, Planck PR4 and ACT lensing maps, together with BAO measurements from DESI DR2, we obtain $H_0 = 69.61^{+0.60}_{-0.55} \rm \, km \, s^{-1} \, Mpc^{-1}$ and $S_8= 0.808\pm0.012$, with $Δm_e / m_e = 0.0109^{+0.0068}_{-0.0066}$ and $A_{\rm lens} = 1.030^{+0.039}_{-0.037}$, both exceeding the $Λ$CDM expectations. We find no indication of spatial curvature deviating from flatness, even when including the Cosmic Chronometers and SNe Ia samples. However, when adopting the latest Planck power spectra likelihoods, NPIPE and HiLLiPoP, we obtain lower electron masses with $Δm_e / m_e = -0.0063^{+0.0095}_{-0.0099}$ and $-0.0095^{+0.0078}_{-0.0079}$, relieving the $S_8$ tension only. The lensing amplitude remains anomalously high, with $A_{\rm lens} = 1.053^{+0.042}_{-0.040}$ and $1.075^{+0.044}_{-0.043}$. Our results point to a promising direction for cosmological models to reconcile the aforementioned discrepancies, although more precise data from future experiments will be necessary to clarify the aforementioned modifications.

astro-ph.CO↗
arXiv2025

Ultra High-Resolution Image Inpainting with Patch-Based Content Consistency Adapter

In this work, we present Patch-Adapter, an effective framework for high-resolution text-guided image inpainting. Unlike existing methods limited to lower resolutions, our approach achieves 4K+ resolution while maintaining precise content consistency and prompt alignment, two critical challenges in image inpainting that intensify with increasing resolution and texture complexity. Patch-Adapter leverages a two-stage adapter architecture to scale the diffusion model's resolution from 1K to 4K+ without requiring structural overhauls: (1) Dual Context Adapter learns coherence between masked and unmasked regions at reduced resolutions to establish global structural consistency; and (2) Reference Patch Adapter implements a patch-level attention mechanism for full-resolution inpainting, preserving local detail fidelity through adaptive feature fusion. This dual-stage architecture uniquely addresses the scalability gap in high-resolution inpainting by decoupling global semantics from localized refinement. Experiments demonstrate that Patch-Adapter not only resolves artifacts common in large-scale inpainting but also achieves state-of-the-art performance on the OpenImages and Photo-Concept-Bucket datasets, outperforming existing methods in both perceptual quality and text-prompt adherence.

cs.CV↗