arXiv subjects

Yinan Zhao

Publications and source records attributed to Yinan Zhao.

At least 19 recordsLinked to original sources

arXiv2026

Wuying-Browser-Agent: Real-World Centric Fundamental Long-Horizon Browser Agents

Browser agents perform well on short, clean demonstrations, but real deployment is fundamentally different: agents must sustain dozens of decisions on live websites while recovering from mistakes and navigating complex UIs. We argue that closing this gap requires alignment at every level of the pipeline, including execution, supervision, optimization, and evaluation, rather than scale alone. We present Wuying-Browser-Agent, a unified framework that addresses each of these levels. A structured browser harness provides stable execution primitives and decision-oriented context management. Reflection and UI-specialized Curriculum SFT (RUIC-SFT) explicitly trains on recovery trajectories and complex-UI interactions. Divergence-Aware Online GRPO (DAO-GRPO) improves long-horizon credit assignment through potential-based reward shaping and divergence-aware step weighting. Finally, we introduce BrowserBench, a bilingual real-web benchmark of 350 tasks averaging 37.9 steps, because most existing benchmarks are too short to expose long-horizon failure modes. Wuying-Browser-Agent-27B achieves 80.6\% on WebVoyager, 66.7\% on Online-Mind2Web, and 65.1\% on BrowserBench, establishing a new open-source state of the art on browser-use benchmarks. The same pipeline also transfers beyond browser use, demonstrating strong general agentic ability and reaching an average score of 73.8 on Tau2-Bench, Claw-Eval, and BFCL-v4.

cs.AI↗

arXiv2026

Inverse-designed meta processing units for multi-task near-field photonic computing

Integrated photonic neural networks require optical operators that are simultaneously compact, matrix-general and compatible with task-level reconfigurability. Here we introduce a meta processing unit (MPU), an inverse-designed near-field photonic device that implements local complex matrix transformations within a shallow-etched silicon region. Each 2x2 operator occupies 9.6 umx4.8 um and is designed as a reusable passive matrix primitive that can be combined with reconfigurable MZI neurons. We demonstrate a 3-bit quantized MZI-equivalent unitary device library with an effective reconstruction precision of 3.32 bits. Beyond unitary operators, we validate arbitrary complex 2x2 matrix fitting and a cascaded 4x4 matrix operation with 92.7% fidelity. We further integrate the MPU with active photonic components and hardware-in-the-loop training, achieving test accuracies of 83.5% and 80.9% on dual-task vowel recognition. In large-scale EMNIST simulations, a fine-grained neuron-level MPU replacement strategy reaches 87.64% average accuracy at 90% shared-MPU replacement, outperforming a layer-level baseline by 7.26 percentage points. These results establish inverse-designed MPUs as compact passive matrix operators for heterogeneous, hardware-adaptive photonic neural networks.

physics.optics↗

arXiv2026

Modeling Doppler Shifts in Radial-Velocity Data with Deep Learning toward Earth-mass Exoplanet Detection

Detecting the tiny Doppler shifts induced by Earth-mass planets in stellar radial-velocity measurements remains extremely challenging due to stellar activity. Many deep-learning methods performing well on simulated data remain difficult to apply reliably on real stellar spectra. The aim of this work is to develop a deep-learning framework that generalizes to real, unseen spectra and improves the detectability of Earth-mass planets in radial-velocity data. We train artificial neural networks on HARPS-N solar spectra with injected planetary signals, using physics-motivated spectral representations based on flux and line-formation temperature, together with their velocity gradients. Two training strategies are explored: hold-out testing and cross-validation. Model robustness is enhanced through genetic-algorithm-based hyperparameter optimization, and predictive uncertainty is quantified using Monte Carlo dropout. Our most precise neural network model reliably retrieves, under the cross-validation strategy, the amplitudes, phases, and orbital periods of planetary signals with amplitudes greater than or equal to 25 cm/s and periods between 10 and 550 days. In addition, in all cases tested here, the successfully recovered signals correspond to the most significant peaks in the periodograms of the Doppler-shift predictions. Temperature-based spectral-shell representations consistently outperform flux-based shells. We also release doppleriann, a Python package implementing the proposed framework. Our results demonstrate that combining physically motivated spectral representations with deep learning provides a promising pathway toward the detection of Earth-mass planets in radial-velocity data from real observations, supported by a modeling framework that is both physically grounded and statistically rigorous, incorporating uncertainty quantification and optimized training strategies.

astro-ph.IM↗

arXiv2025

MapReduce LoRA: Advancing the Pareto Front in Multi-Preference Optimization for Generative Models

Reinforcement learning from human feedback (RLHF) with reward models has advanced alignment of generative models to human aesthetic and perceptual preferences. However, jointly optimizing multiple rewards often incurs an alignment tax, improving one dimension while degrading others. To address this, we introduce two complementary methods: MapReduce LoRA and Reward-aware Token Embedding (RaTE). MapReduce LoRA trains preference-specific LoRA experts in parallel and iteratively merges them to refine a shared base model; RaTE learns reward-specific token embeddings that compose at inference for flexible preference control. Experiments on Text-to-Image generation (Stable Diffusion 3.5 Medium and FLUX.1-dev) show improvements of 36.1%, 4.6%, and 55.7%, and 32.7%, 4.3%, and 67.1% on GenEval, PickScore, and OCR, respectively. On Text-to-Video generation (HunyuanVideo), visual and motion quality improve by 48.1% and 90.0%, respectively. On the language task, Helpful Assistant, with Llama-2 7B, helpful and harmless improve by 43.4% and 136.7%, respectively. Our framework sets a new state-of-the-art multi-preference alignment recipe across modalities.

cs.CV↗

arXiv2025

The validity of the universal relation between corrections to entropy and the extremality of Schwarzschild-de Sitter black holes under the GUP and EUP

We investigate the extremality relations by examining perturbative corrections to both the entropy of Schwarzschild-de Sitter black holes and their extremality bounds under the Nariai limit within the frame of the generalized uncertainty principle (GUP) and the extended uncertainty principle (EUP) respectively. The GUP-corrected or EUP-corrected horizons bring about the additional terms finally in the Goon-Penco relation to violate the universal ones and these extra terms cannot be canceled by adding the adequate conditions. We argue that the corrected uncertainty principles including GUP and EUP violate the validity of extremality relations because no matching condition can be imposed to support the Goon-Penco relation unless the influences from GUP and EUP disappear.

hep-th↗

arXiv2025

The thermodynamic stability and phase structure of the Einstein-Euler-Heisenberg-AdS black holes

In both canonical ensemble and grand canonical ensemble, the thermodynamic stability and phase structure of Einstein-Euler-Heisenberg-AdS black hole are studied. We derive the Hawking temperature, Helmholtz free energy, Gibbs potential, entropy and heat capacity of the black holes. We compute the minimum temperature to find that the phase transition may happen at the lowest point. The entropy-temperature diagram consists of two parts. The upper part belonging to the large black holes under the influence from the electromagnetic self-interactions keeps the positive heat capacity, leading the huge compact objects to survive. The lower curves corresponding to the small ones show that the heat capacity of the tiny black holes is negative, which means that the nonlinear-effect-corrected smaller sources will evaporate. The further discussions show that the nonlinear effect modifies the thermodynamic quantities, but the corrections limited by the nonlinear factor $μ$ with allowed values can not change the properties and the phase structure fundamentally and thoroughly. We argue that the influence from self-interaction can not make the Einstein-Euler-Heisenberg-AdS black holes to split under the second law of thermodynamics.

hep-th↗

arXiv2024

Precise and efficient modeling of stellar-activity-affected solar spectra using SOAP-GPU

One of the main obstacles in exoplanet detection when using the radial velocity (RV) technique is the presence of stellar activity signal induced by magnetic regions. In this context, a realistic simulated dataset that can provide photometry and spectroscopic outputs is needed for method development. The goal of this paper is to describe two realistic simulations of solar activity obtained from SOAP-GPU and to compare them with real data obtained from the HARPS-N solar telescope. We describe two different methods of modeling solar activity using SOAP-GPU. The first models the evolution of active regions based on the spot number as a function of time. The second method relies on the extraction of active regions from the Solar Dynamics Observatory (SDO) data. The simulated spectral time series generated with the first method shows a long-term RV behavior similar to that seen in the HARPS-N solar observations. The effect of stellar activity induced by stellar rotation is also well modeled with prominent periodicities at the stellar rotation period and its first harmonic. The comparison between the simulated spectral time series generated using SDO images and the HARPS-N solar spectra shows that SOAP-GPU can precisely model the RV time series of the Sun to a precision better than 0.9 m/s. By studying the width and depth variations of each spectral line in the HARPS-N solar and SOAP-GPU data, we find a strong correlation between the observation and the simulation for strong spectral lines, therefore supporting the modeling of the stellar activity effect at the spectral level. These simulated solar spectral time series serve as a useful test bed for evaluating spectral-level stellar activity mitigation techniques.

astro-ph.SR↗

arXiv2024

Discovery of Small Ultra-short-period Planets Orbiting KG Dwarfs in Kepler Survey Using GPU Phase Folding and Deep Learning Detection System

Of over 5,000 exoplanets identified so far, only a few hundred possess sub-Earth radii. The formation processes of these sub-Earths remain elusive, and acquiring additional samples is essential for investigating this unique population. In our study, we employ the GPFC method, a novel GPU Phase Folding algorithm combined with a Convolutional Neural Network, on Kepler photometry data. This method enhances the transit search speed significantly over the traditional Box-fitting Least Squares method, allowing a complete search of the known Kepler KOI data within days using a commercial GPU card. To date, we have identified five new ultra-short-period planets (USPs): Kepler-158d, Kepler-963c, Kepler-879c, Kepler-1489c, and KOI-4978.02. Kepler-879c with a radius of $0.4 R_\oplus$ completes its orbit around a G dwarf in 0.646716 days. Kepler-158d with a radius of $0.43 R_\oplus$ orbits a K dwarf star every 0.645088 days. Kepler-1489c with a radius of $0.51 R_\oplus$ orbits a G dwarf in 0.680741 days. Kepler-963c with a radius of $0.6 R_\oplus$ revolves around a G dwarf in 0.919783 days, and KOI-4978.02 with a radius of $0.7 R_\oplus$ circles a G dwarf in 0.941967 days. Among our findings, Kepler-879c, Kepler-158d and Kepler-963c rank as the first, the third, the fourth smallest USPs identified to date. Notably, Kepler-158d stands as the smallest USP found orbiting K dwarfs while Kepler-963c, Kepler-879c, Kepler-1489c, and KOI-4978.02 are the smallest USPs found orbiting G dwarfs. Kepler-879c, Kepler-158d, Kepler-1489c, and KOI-4978.02 are among the smallest planets that are closest to their host stars, with orbits within 5 stellar radii. In addition, these discoveries highlight GPFC's promising capability in identifying small, new transiting exoplanets within photometry data from Kepler, TESS, and upcoming space transit missions, PLATO and ET.

astro-ph.EP↗

arXiv2024

Improving Earth-like planet detection in radial velocity using deep learning

Many novel methods have been proposed to mitigate stellar activity for exoplanet detection as the presence of stellar activity in radial velocity (RV) measurements is the current major limitation. Unlike traditional methods that model stellar activity in the RV domain, more methods are moving in the direction of disentangling stellar activity at the spectral level. The goal of this paper is to present a novel convolutional neural network-based algorithm that efficiently models stellar activity signals at the spectral level, enhancing the detection of Earth-like planets. We trained a convolutional neural network to build the correlation between the change in the spectral line profile and the corresponding RV, full width at half maximum (FWHM) and bisector span (BIS) values derived from the classical cross-correlation function. This algorithm has been tested on three intensively observed stars: Alpha Centauri B (HD128621), Tau ceti (HD10700), and the Sun. By injecting simulated planetary signals at the spectral level, we demonstrate that our machine learning algorithm can achieve, for HD128621 and HD10700, a detection threshold of 0.5 m/s in semi-amplitude for planets with periods ranging from 10 to 300 days. This threshold would correspond to the detection of a $\sim$4$\mathrm{M}_{\oplus}$ in the habitable zone of those stars. On the HARPS-N solar dataset, our algorithm is even more efficient at mitigating stellar activity signals and can reach a threshold of 0.2 m/s, which would correspond to a 2.2$\mathrm{M}_{\oplus}$ planet on the orbit of the Earth. To the best of our knowledge, it is the first time that such low detection thresholds are reported for the Sun, but also for other stars, and therefore this highlights the efficiency of our convolutional neural network-based algorithm at mitigating stellar activity in RV measurements.

astro-ph.EP↗

arXiv2024

AVID: Any-Length Video Inpainting with Diffusion Model

Recent advances in diffusion models have successfully enabled text-guided image inpainting. While it seems straightforward to extend such editing capability into the video domain, there have been fewer works regarding text-guided video inpainting. Given a video, a masked region at its initial frame, and an editing prompt, it requires a model to do infilling at each frame following the editing guidance while keeping the out-of-mask region intact. There are three main challenges in text-guided video inpainting: ($i$) temporal consistency of the edited video, ($ii$) supporting different inpainting types at different structural fidelity levels, and ($iii$) dealing with variable video length. To address these challenges, we introduce Any-Length Video Inpainting with Diffusion Model, dubbed as AVID. At its core, our model is equipped with effective motion modules and adjustable structure guidance, for fixed-length video inpainting. Building on top of that, we propose a novel Temporal MultiDiffusion sampling pipeline with a middle-frame attention guidance mechanism, facilitating the generation of videos with any desired duration. Our comprehensive experiments show our model can robustly deal with various inpainting types at different video duration ranges, with high quality. More visualization results are made publicly available at https://zhang-zx.github.io/AVID/ .

cs.CV↗

arXiv2024

The GPU Phase Folding and Deep Learning Method for Detecting Exoplanet Transits

This paper presents GPFC, a novel Graphics Processing Unit (GPU) Phase Folding and Convolutional Neural Network (CNN) system to detect exoplanets using the transit method. We devise a fast folding algorithm parallelized on a GPU to amplify low signal-to-noise ratio transit signals, allowing a search at high precision and speed. A CNN trained on two million synthetic light curves reports a score indicating the likelihood of a planetary signal at each period. While the GPFC method has broad applicability across period ranges, this research specifically focuses on detecting ultra-short-period planets with orbital periods less than one day. GPFC improves on speed by three orders of magnitude over the predominant Box-fitting Least Squares (BLS) method. Our simulation results show GPFC achieves $97%$ training accuracy, higher true positive rate at the same false positive rate of detection, and higher precision at the same recall rate when compared to BLS. GPFC recovers $100\%$ of known ultra-short-period planets in $\textit{Kepler}$ light curves from a blind search. These results highlight the promise of GPFC as an alternative approach to the traditional BLS algorithm for finding new transiting exoplanets in data taken with $\textit{Kepler}$ and other space transit missions such as K2, TESS and future PLATO and Earth 2.0.

astro-ph.EP↗

arXiv2023

FlowVid: Taming Imperfect Optical Flows for Consistent Video-to-Video Synthesis

Diffusion models have transformed the image-to-image (I2I) synthesis and are now permeating into videos. However, the advancement of video-to-video (V2V) synthesis has been hampered by the challenge of maintaining temporal consistency across video frames. This paper proposes a consistent V2V synthesis framework by jointly leveraging spatial conditions and temporal optical flow clues within the source video. Contrary to prior methods that strictly adhere to optical flow, our approach harnesses its benefits while handling the imperfection in flow estimation. We encode the optical flow via warping from the first frame and serve it as a supplementary reference in the diffusion model. This enables our model for video synthesis by editing the first frame with any prevalent I2I models and then propagating edits to successive frames. Our V2V model, FlowVid, demonstrates remarkable properties: (1) Flexibility: FlowVid works seamlessly with existing I2I models, facilitating various modifications, including stylization, object swaps, and local edits. (2) Efficiency: Generation of a 4-second video with 30 FPS and 512x512 resolution takes only 1.5 minutes, which is 3.1x, 7.2x, and 10.5x faster than CoDeF, Rerender, and TokenFlow, respectively. (3) High-quality: In user studies, our FlowVid is preferred 45.7% of the time, outperforming CoDeF (3.5%), Rerender (10.2%), and TokenFlow (40.4%).

cs.CV↗

arXiv2023

Open-Vocabulary Semantic Segmentation with Mask-adapted CLIP

Open-vocabulary semantic segmentation aims to segment an image into semantic regions according to text descriptions, which may not have been seen during training. Recent two-stage methods first generate class-agnostic mask proposals and then leverage pre-trained vision-language models, e.g., CLIP, to classify masked regions. We identify the performance bottleneck of this paradigm to be the pre-trained CLIP model, since it does not perform well on masked images. To address this, we propose to finetune CLIP on a collection of masked image regions and their corresponding text descriptions. We collect training data by mining an existing image-caption dataset (e.g., COCO Captions), using CLIP to match masked image regions to nouns in the image captions. Compared with the more precise and manually annotated segmentation labels with fixed classes (e.g., COCO-Stuff), we find our noisy but diverse dataset can better retain CLIP's generalization ability. Along with finetuning the entire model, we utilize the "blank" areas in masked images using a method we dub mask prompt tuning. Experiments demonstrate mask prompt tuning brings significant improvement without modifying any weights of CLIP, and it can further improve a fully finetuned model. In particular, when trained on COCO and evaluated on ADE20K-150, our best model achieves 29.6% mIoU, which is +8.5% higher than the previous state-of-the-art. For the first time, open-vocabulary generalist models match the performance of supervised specialist models in 2017 without dataset-specific adaptations.

cs.CV↗

arXiv2023

SOAP-GPU: Efficient Spectral Modelling of Stellar Activity Using Graphical Processing Units

Stellar activity mitigation is one of the major challenges for the detection of earth-like exoplanets in radial velocity (RV) measurements. Several promising techniques are now investigating the use of spectral time-series, to differentiate between stellar and planetary perturbations. In this paper, we present a new version of the Spot Oscillation And Planet (SOAP) 2.0 code that can model stellar activity at the spectral level using graphical processing units (GPUs). We take advantage of the computational power of GPUs to optimise the computationally expensive algorithms behind the original SOAP 2.0 code. We develope GPU kernels that allow to model stellar activity on any given wavelength range. In addition to the treatment of stellar activity at the spectral level, SOAP-GPU also includes the change of spectral line bisectors from center to limb, and can take as input PHOENIX spectra to model the quiet photosphere, spots and faculae, which allow to simulate stellar activity for a wide space in stellar properties. Benchmark calculations show that for the same accuracy, this new code improves the computational speed by a factor of 60 compared with a modified version of SOAP 2.0 that generates spectra, when modeling stellar activity on the full visible spectral range with a resolution of R=115'000. Although the code now includes the variation of spectral line bisector with center to limb angle, the effect on the derived RVs is small. The publicly available SOAP-GPU code allows to efficiently model stellar activity at the spectral level, which is essential to test further stellar activity mitigation techniques working at the level of spectral timeseries not affected by other sources of noise. Besides a huge gain in performance, SOAP-GPU also includes more physics and is able to model different stars than the Sun, from F to K dwarfs, thanks to the use of the PHOENIX spectral library.

astro-ph.SR↗

arXiv2022

Newly discovered Ca II absorbers in the early universe: statistics, element abundances and dust

We report discoveries of 165 new quasar Ca II absorbers from the Sloan Digital Sky Survey (SDSS) Data Release 7 and 12. Our Ca II rest frame equivalent width distribution supports the weak and strong subpopulations, split at ${W}^{\lambda3934}_{0}=0.7$Å. Comparison of both populations' dust depletion shows clear consistency for weak absorber association with halo-type gas in the Milky Way (MW) while strong absorbers have environments consistent with halo and disc-type gas. We probed our high redshift Ca II absorbers for 2175Å dust bumps, discovering 12 2175Å dust absorbers (2DAs). This clearly shows that some Ca II absorbers follow the Large Magellanic Cloud (LMC) extinction law rather than the Small Magellanic Cloud extinction law. About 33% of our strong Ca II absorbers exhibit the 2175Å dust bump while only 6% of weak Ca II absorbers show this bump. 2DA detection further supports the theory that strong Ca II absorbers are associated with disk components and are dustier than the weak population. Comparing average Ca II absorber dust depletion patterns to that of Damped Lyα Absorbers (DLAs), Mg II absorbers, and 2DAs shows that Ca II absorbers generally have environments with more dust than DLAs and Mg II absorbers, but less dust than 2DAs. Comparing 2175Å dust bump strengths from different samples and also the MW and LMC, the bump strength appears to grow stronger as the redshift decreases, indicating dust growth and the global chemical enrichment of galaxies in the universe over time.

astro-ph.GA↗

arXiv2022

Discovering Ca II Absorption Lines With a Neural Network

Quasar absorption line analysis is critical for studying gas and dust components and their physical and chemical properties as well as the evolution and formation of galaxies in the early universe. Ca II absorbers, which are one of the dustiest absorbers and are located at lower redshifts than most other absorbers, are especially valuable when studying physical processes and conditions in recent galaxies. However, the number of known quasar Ca II absorbers is relatively low due to the difficulty of detecting them with traditional methods. In this work, we developed an accurate and quick approach to search for Ca II absorption lines using deep learning. In our deep learning model, a convolutional neural network, tuned using simulated data, is used for the classification task. The simulated training data are generated by inserting artificial Ca II absorption lines into original quasar spectra from the Sloan Digital Sky Survey (SDSS) whilst an existing Ca II catalog is adopted as the test set. The resulting model achieves an accuracy of 96% on the real data in the test set. Our solution runs thousands of times faster than traditional methods, taking a fraction of a second to analyze thousands of quasars while traditional methods may take days to weeks. The trained neural network is applied to quasar spectra from SDSS's DR7 and DR12 and discovered 399 new quasar Ca II absorbers. In addition, we confirmed 409 known quasar Ca II absorbers identified previously by other research groups through traditional methods.

astro-ph.GA↗

arXiv2022

ET White Paper: To Find the First Earth 2.0

We propose to develop a wide-field and ultra-high-precision photometric survey mission, temporarily named "Earth 2.0 (ET)". This mission is designed to measure, for the first time, the occurrence rate and the orbital distributions of Earth-sized planets. ET consists of seven 30cm telescopes, to be launched to the Earth-Sun's L2 point. Six of these are transit telescopes with a field of view of 500 square degrees. Staring in the direction that encompasses the original Kepler field for four continuous years, this monitoring will return tens of thousands of transiting planets, including the elusive Earth twins orbiting solar-type stars. The seventh telescope is a 30cm microlensing telescope that will monitor an area of 4 square degrees toward the galactic bulge. This, combined with simultaneous ground-based KMTNet observations, will measure masses for hundreds of long-period and free-floating planets. Together, the transit and the microlensing telescopes will revolutionize our understandings of terrestrial planets across a large swath of orbital distances and free space. In addition, the survey data will also facilitate studies in the fields of asteroseismology, Galactic archeology, time-domain sciences, and black holes in binaries.

astro-ph.IM↗

arXiv2022

Human Behavior Recognition Method Based on CEEMD-ES Radar Selection

In recent years, the millimeter-wave radar to identify human behavior has been widely used in medical,security, and other fields. When multiple radars are performing detection tasks, the validity of the features contained in each radar is difficult to guarantee. In addition, processing multiple radar data also requires a lot of time and computational cost. The Complementary Ensemble Empirical Mode Decomposition-Energy Slice (CEEMD-ES) multistatic radar selection method is proposed to solve these problems. First, this method decomposes and reconstructs the radar signal according to the difference in the reflected echo frequency between the limbs and the trunk of the human body. Then, the radar is selected according to the difference between the ratio of echo energy of limbs and trunk and the theoretical value. The time domain, frequency domain and various entropy features of the selected radar are extracted. Finally, the Extreme Learning Machine (ELM) recognition model of the ReLu core is established. Experiments show that this method can effectively select the radar, and the recognition rate of three kinds of human actions is 98.53%.

eess.SP↗