arXiv subjects

Jinglin Zhao

Publications and source records attributed to Jinglin Zhao.

18 recordsLinked to original sources

arXiv2025

VisChainBench: A Benchmark for Multi-Turn, Multi-Image Visual Reasoning Beyond Language Priors

Understanding multi-image, multi-turn scenarios is a critical yet underexplored capability for Large Vision-Language Models (LVLMs). Existing benchmarks predominantly focus on static or horizontal comparisons -- e.g., spotting visual differences or assessing appropriateness -- while relying heavily on language cues. Such settings overlook progressive, context-dependent reasoning and the challenge of visual-to-visual inference. To bridge this gap, we present VisChainBench, a large-scale benchmark designed to rigorously evaluate LVLMs' ability to perform multi-step visual reasoning across sequential, interdependent tasks with minimal language guidance. VisChainBench contains 1,457 tasks spanning over 20,000 images across three diverse domains (e.g., daily scenarios, engineering troubleshooting), structured to mimic real-world decision-making processes. Uniquely, the benchmark is constructed using a multi-agent generation pipeline, ensuring high visual diversity and controlled language bias. All the benchmark data and code for benchmark construction are available for viewing and download via following Link: https://huggingface.co/datasets/eyehole/VisChainBench

cs.CV↗

arXiv2025

Time-resolved p-mode oscillations for subgiant HD 142091 with NEID at WIYN

Detections of Earth-analog planets in radial velocity observations are limited by stellar astrophysical variability occurring on a variety of timescales. Current state-of-the-art methods to disentangle potential planet signals from intrinsic stellar signals assume that stellar signals introduce asymmetries to the line profiles that can therefore be separated from the pure translational Doppler shifts of planets. Here, we examine this assumption using a time series of resolved stellar p-mode oscillations in HD 142091 ($\kappa$ CrB), as observed on a single night with the NEID spectrograph at 2-minute cadence and with 25 cm/s precision. As an evolved subgiant star, this target has p-mode oscillations that are larger in amplitude (4-8 m/s) and occur on longer timescales (80 min.) than those of typical Sun-like stars of RV surveys, magnifying their corresponding effects on the stellar spectral profile. We show that for HD 142091, p-mode oscillations manifest primarily as pure Doppler shifts in the average line profile -- measured by the cross-correlation function (CCF) -- with "shape-driven" CCF variations as a higher-order effect. Specifically, we find that the amplitude of the shift varies across the CCF bisector, with 10% larger oscillation amplitudes closer to the core of the CCF, and 25% smaller oscillation amplitudes for bisector velocities derived near the wings; we attribute this trend to larger oscillation velocities higher in the stellar atmosphere. Using a line-by-line analysis, we verify that a similar trend is seen as a function of average line depth, with deeper lines showing larger oscillation amplitudes. Finally, we find no evidence that p-mode oscillations have a chromatic dependence across the NEID bandpass beyond that due to intrinsic line depth differences across the spectrum.

astro-ph.SR↗

arXiv2024

Polarization position angle standard stars: a reassessment of $\theta$ and its variability for seventeen stars based on a decade of observations

Observations of polarization position angle ($\theta$) standards made from 2014 to 2023 with the High Precision Polarimetric Instrument (HIPPI) and other HIPPI-class polarimeters in both hemispheres are used to investigate their variability. Multi-band data were first used to thoroughly recalibrate the instrument performance by bench-marking against carefully selected literature data. A novel Co-ordinate Difference Matrix (CDM) approach - which combines pairs of points - was then used to amalgamate monochromatic ($g^\prime$ band) observations from many observing runs and re-determine $\theta$ for 17 standard stars. The CDM algorithm was then integrated into a fitting routine and used to establish the impact of stellar variability on the measured position angle scatter. The approach yields variability detections for stars on long time scales that appear stable over short runs. The best position angle standards are $\ell$ Car, $o$ Sco, HD 154445, HD 161056 and $\iota^1$ Sco which are stable to $\leq$ 0.123$^\circ$. Position angle variability of 0.27-0.82$^\circ$, significant at the 3-$\sigma$ level, is found for 5 standards, including the Luminous Blue Variable HD 160529 and all but one of the other B/A-type supergiants (HD 80558, HD 111613, HD 183143 and 55 Cyg), most of which also appear likely to be variable in polarization magnitude ($p$) - there is no preferred orientation for the polarization in these objects, which are all classified as $\alpha$ Cygni variables. Despite this we make six key recommendations for observers - relating to data acquisition, processing and reporting - that will allow them to use these standards to achieve $<$ 0.1$^\circ$ precision in the telescope position angle with similar instrumentation, and allow data sets to be combined more accurately.

astro-ph.SR↗

arXiv2024

Earths within Reach: Evaluation of Strategies for Mitigating Solar Variability using 3.5 years of NEID Sun-as-a-Star Observations

We present the results of Sun-as-a-star observations by the NEID Solar Telescope at WIYN Observatory, spanning January 1, 2021 through June 30, 2024. We identify 117,060 observations which are unlikely to be significantly affected by weather, hardware or major calibration issues. We describe several high-level data products being made available to the community to aid in the interpretation and inter comparisons of NEID solar observations. Solar observations demonstrate excellent performance of NEID, including radial velocity (RV) accuracy and long-term stability of better than $\simeq 0.37$ m s$^{-1}$ over $\simeq 3.5$ years, even though NEID was not originally designed or optimized for daytime observations of the Sun. Currently, intrinsic stellar variability is the primary barrier to detecting Earth-analog planets for most nearby, Sun-like stars. We present a comparison of the effectiveness of several methods proposed to mitigate the effects of solar variability on the Sun's estimated RV. We find that the Scalpels algorithm performs particularly well and substantially reduces the RMS RV of solar spectra from over 2 m s$^{-1}$ to 0.277 m s$^{-1}$. Even when training on a subset of days with NEID solar observations and testing on a held-out sample, the RMS of cleaned RV is 0.34-0.42 m s$^{-1}$. This is significantly better than previous attempts at removing solar variability and suggests that the current generation of EPRV instruments are technically capable of detecting Earth-mass planets orbiting a solar twin if provided with sufficient observing time allocations ($\sim 10^3$ nights of observations).

astro-ph.SR↗

arXiv2024

DTN: Deep Multiple Task-specific Feature Interactions Network for Multi-Task Recommendation

Neural-based multi-task learning (MTL) has been successfully applied to many recommendation applications. However, these MTL models (e.g., MMoE, PLE) did not consider feature interaction during the optimization, which is crucial for capturing complex high-order features and has been widely used in ranking models for real-world recommender systems. Moreover, through feature importance analysis across various tasks in MTL, we have observed an interesting divergence phenomenon that the same feature can have significantly different importance across different tasks in MTL. To address these issues, we propose Deep Multiple Task-specific Feature Interactions Network (DTN) with a novel model structure design. DTN introduces multiple diversified task-specific feature interaction methods and task-sensitive network in MTL networks, enabling the model to learn task-specific diversified feature interaction representations, which improves the efficiency of joint representation learning in a general setup. We applied DTN to our company's real-world E-commerce recommendation dataset, which consisted of over 6.3 billion samples, the results demonstrated that DTN significantly outperformed state-of-the-art MTL models. Moreover, during online evaluation of DTN in a large-scale E-commerce recommender system, we observed a 3.28% in clicks, a 3.10% increase in orders and a 2.70% increase in GMV (Gross Merchandise Value) compared to the state-of-the-art MTL models. Finally, extensive offline experiments conducted on public benchmark datasets demonstrate that DTN can be applied to various scenarios beyond recommendations, enhancing the performance of ranking models.

cs.IR↗

arXiv2024

The death of Vulcan: NEID reveals the planet candidate orbiting HD 26965 is stellar activity

We revisit the long-studied radial velocity (RV) target HD26965 using recent observations from the NASA-NSF 'NEID' precision Doppler facility. Leveraging a suite of classical activity indicators, combined with line-by-line RV analyses, we demonstrate that the claimed 45-day signal previously identified as a planet candidate is most likely an activity-induced signal. Correlating the bulk (spectrally-averaged) RV with canonical line activity indicators confirms a multi-day 'lag' between the observed activity indicator time series and the measured RV. When accounting for this lag, we show that much of the observed RV signal can be removed by a linear detrending of the data. Investigating activity at the line-by-line level, we find a depth-dependent correlation between individual line RVs and the bulk RVs, further indicative of periodic suppression of convective blueshift causing the observed RV variability, rather than an orbiting planet. We conclude that the combined evidence of the activity correlations and depth dependence is consistent with a radial velocity signature dominated by a rotationally-modulated activity signal at a period of $\sim$42 days. We hypothesize that this activity signature is due to a combination of spots and convective blueshift suppression. The tools applied in our analysis are broadly applicable to other stars, and could help paint a more comprehensive picture of the manifestations of stellar activity in future Doppler RV surveys.

astro-ph.EP↗

arXiv2024

Mitigating Prior Shape Bias in Point Clouds via Differentiable Center Learning

Masked autoencoding and generative pretraining have achieved remarkable success in computer vision and natural language processing, and more recently, they have been extended to the point cloud domain. Nevertheless, existing point cloud models suffer from the issue of information leakage due to the pre-sampling of center points, which leads to trivial proxy tasks for the models. These approaches primarily focus on local feature reconstruction, limiting their ability to capture global patterns within point clouds. In this paper, we argue that the reduced difficulty of pretext tasks hampers the model's capacity to learn expressive representations. To address these limitations, we introduce a novel solution called the Differentiable Center Sampling Network (DCS-Net). It tackles the information leakage problem by incorporating both global feature reconstruction and local feature reconstruction as non-trivial proxy tasks, enabling simultaneous learning of both the global and local patterns within point cloud. Experimental results demonstrate that our method enhances the expressive capacity of existing point cloud models and effectively addresses the issue of information leakage.

cs.CV↗

arXiv2024

Revised Architecture and Two New Super-Earths in the HD 134606 Planetary System

Multi-planet systems exhibit a diversity of architectures that diverge from the solar system and contribute to the topic of exoplanet demographics. Radial velocity (RV) surveys form a crucial component of exoplanet surveys, as their long observational baselines allow searches for more distant planetary orbits. This work provides a significantly revised architecture for the multi-planet system HD 134606 using both HARPS and UCLES RVs. We confirm the presence of previously reported planets b, c, and d with periods $12.0897^{+0.0019}_{-0.0018}$, $58.947^{+0.056}_{-0.054}$, and $958.7^{+6.3}_{-5.9}$ days, and masses $9.14^{+0.65}_{-0.63}$, $11.0\pm1$, and $44.5\pm2.9$ Earth masses respectively, with the planet d orbit significantly revised to over double that originally reported. We report two newly detected super-Earths, e and f, with periods $4.31943^{+0.00075}_{-0.00068}$ and $26.9^{+0.019}_{-0.017}$ days, and masses $2.31^{+0.36}_{-0.35}$ and $5.52^{+0.74}_{-0.73}$ Earth masses, respectively. In addition, we identify a linear trend in the RV time series, and the cause of this acceleration is deemed to be a newly detected sub-stellar companion at large separation. HD 134606 now displays four low mass planets in a compact region near the star, one gas giant further out in the Habitable Zone, an additional massive companion in the outer regime, and a low mass M dwarf stellar companion at large separation, making it an intriguing target for system formation/evolution studies. The location of planet d in the Habitable Zone proves to be an exciting candidate for future space-based direct imaging missions, whereas continued RV observations of this system are recommended for understanding the nature of the massive, long period companion.

astro-ph.EP↗

arXiv2023

Doppler Constraints on Planetary Companions to Nearby Sun-like Stars: An Archival Radial Velocity Survey of Southern Targets for Proposed NASA Direct Imaging Missions

Directly imaging temperate rocky planets orbiting nearby, Sun-like stars with a 6-m-class IR/O/UV space telescope, recently dubbed the Habitable Worlds Observatory, is a high priority goal of the Astro2020 Decadal Survey. To prepare for future direct imaging surveys, the list of potential targets should be thoroughly vetted to maximize efficiency and scientific yield. We present an analysis of archival radial velocity data for southern stars from the NASA/NSF Extreme Precision Radial Velocity Working Group's list of high priority target stars for future direct imaging missions (drawn from the HabEx, LUVOIR, and Starshade studies). For each star, we constrain the region of companion mass and period parameter space we are already sensitive to based on the observational baseline, sampling, and precision of the archival RV data. Additionally, for some of the targets we report new estimates of magnetic activity cycle periods, rotation periods, improved orbital parameters for previously known exoplanets, and new candidate planet signals that require further vetting or observations to confirm. Our results show that for many of these stars we are not yet sensitive to even Saturn-mass planets in the habitable zone, let alone smaller planets, highlighting the need for future EPRV vetting efforts before the launch of a direct imaging mission. We present evidence that the candidate temperate super-Earth exoplanet HD 85512 b is most likely due to the star's rotation, and report an RV acceleration for delta Pav which supports the existence of a distant giant planet previously inferred from astrometry.

astro-ph.EP↗

arXiv2022

FIESTA II. Disentangling stellar and instrumental variability from exoplanetary Doppler shifts in Fourier domain

The radial velocity (RV) detection of exoplanets is challenged by stellar spectroscopic variability that can mimic the presence of planets and by instrumental instability that can further obscure the detection. Both stellar and instrumental changes can distort the spectral line profiles and be misinterpreted as apparent RV shifts. We present an improved FourIEr \textit{phase} SpecTrum Analysis (FIESTA a.k.a. $\mathitϕ$ESTA) to disentangle apparent velocity shifts due to a line deformation from a true Doppler shift. $\mathitϕ$ESTA projects stellar spectrum's cross correlation function (CCF) onto a truncated set of Fourier basis functions. Using the amplitude and phase information from each Fourier mode, we can trace the line variability at different CCF width scales to robustly identify and mitigate multiple sources of RV contamination. For example, in our study of the 3 years of HARPS-N solar data, $\mathitϕ$ESTA reveals the solar rotational effect, the long-term trend due to solar magnetic cycle, instrumental instability and apparent solar rotation rate changes. Applying a multiple linear regression model on $\mathitϕ$ESTA metrics, we reduce the weighted rms noise from 1.89 m/s to 0.98 m/s. In addition, we observe a $\sim$3 days lag in the $\mathitϕ$ESTA metrics, similar to the findings from previous studies on BIS and FWHM.

astro-ph.EP↗

arXiv2022

The EXPRES Stellar Signals Project II. State of the Field in Disentangling Photospheric Velocities

Measured spectral shifts due to intrinsic stellar variability (e.g., pulsations, granulation) and activity (e.g., spots, plages) are the largest source of error for extreme precision radial velocity (EPRV) exoplanet detection. Several methods are designed to disentangle stellar signals from true center-of-mass shifts due to planets. The EXPRES Stellar Signals Project (ESSP) presents a self-consistent comparison of 22 different methods tested on the same extreme-precision spectroscopic data from EXPRES. Methods derived new activity indicators, constructed models for mapping an indicator to the needed RV correction, or separated out shape- and shift-driven RV components. Since no ground truth is known when using real data, relative method performance is assessed using the total and nightly scatter of returned RVs and agreement between the results of different methods. Nearly all submitted methods return a lower RV RMS than classic linear decorrelation, but no method is yet consistently reducing the RV RMS to sub-meter-per-second levels. There is a concerning lack of agreement between the RVs returned by different methods. These results suggest that continued progress in this field necessitates increased interpretability of methods, high-cadence data to capture stellar signals at all timescales, and continued tests like the ESSP using consistent data sets with more advanced metrics for method performance. Future comparisons should make use of various well-characterized data sets -- such as solar data or data with known injected planetary and/or stellar signals -- to better understand method performance and whether planetary signals are preserved.

astro-ph.EP↗

arXiv2021

Polarization of hot Jupiter systems: a likely detection of stellar activity and a possible detection of planetary polarization

We present high-precision linear polarization observations of four bright hot Jupiter systems ($τ$ Boo, HD 179949, HD 189733 and 51 Peg) and use the data to search for polarized reflected light from the planets. The data for 51 Peg are consistent with a reflected light polarization signal at about the level expected with 2.8$σ$ significance and a false alarm probability of 1.9 per cent. More data will be needed to confirm a detection of reflected light in this system. HD 189733 shows highly variable polarization that appears to be most likely the result of magnetic activity of the host star. This masks any polarization due to reflected light, but a polarization signal at the expected level of $\sim$20 ppm cannot be ruled out. $τ$ Boo and HD 179949 show no evidence for polarization due to reflected light. The results are consistent with the idea that many hot Jupiters have low geometric albedos. Conclusive detection of polarized reflected light from hot Jupiters is likely to require further improvements in instrument sensitivity.

astro-ph.EP↗

arXiv2020

A Topological Nomenclature for 3D Shape Analysis in Connectomics

One of the essential tasks in connectomics is the morphology analysis of neurons and organelles like mitochondria to shed light on their biological properties. However, these biological objects often have tangled parts or complex branching patterns, which make it hard to abstract, categorize, and manipulate their morphology. In this paper, we develop a novel topological nomenclature system to name these objects like the appellation for chemical compounds to promote neuroscience analysis based on their skeletal structures. We first convert the volumetric representation into the topology-preserving reduced graph to untangle the objects. Next, we develop nomenclature rules for pyramidal neurons and mitochondria from the reduced graph and finally learn the feature embedding for shape manipulation. In ablation studies, we quantitatively show that graphs generated by our proposed method align with the perception of experts. On 3D shape retrieval and decomposition tasks, we qualitatively demonstrate that the encoded topological nomenclature features achieve better results than state-of-the-art shape descriptors. To advance neuroscience, we will release a 3D segmentation dataset of mitochondria and pyramidal neurons reconstructed from a 100um cube electron microscopy volume with their reduced graph and topological nomenclature annotations. Code is publicly available at https://github.com/donglaiw/ibexHelper.

cs.CV↗

arXiv2019

The Rotationally Modulated Polarization of $ξ$ Boo A

We have observed the active star $ξ$ Boo A (HD 131156A) with high precision broadband linear polarimetry contemporaneously with circular spectropolarimetry. We find both signals are modulated by the 6.43 day rotation period of $ξ$ Boo A. The signals from the two techniques are 0.25 out of phase, consistent with the broadband linear polarization resulting from differential saturation of spectral lines in the global transverse magnetic field. The mean magnitude of the linear polarization signal is ~4 ppm/G but its structure is complex and the amplitude of the variations suppressed relative to the longitudinal magnetic field. The result has important implications for current attempts to detect polarized light from hot Jupiters orbiting active stars in the combined light of the star and planet. In such work stellar activity will manifest as noise, both on the time scale of stellar rotation, and on longer time scales - where changes in activity level will manifest as a baseline shift between observing runs.

astro-ph.SR↗

arXiv2019

Truly eccentric. I. Revisiting eight single-eccentric planetary systems

We examine eight known single-eccentric planetary systems in light of recently released large data archives and new analysis techniques. For four of these systems (HD 7449, HD 65216, HD 89744, HD 92788) we find evidence for additional long-period companions. HD 65216c is a Jupiter analog, with a period of 14.7 yr, $e=0.18$, and m sin $i$ of 2M_Jup, whilst the remaining candidate companions move on as-yet-incomplete orbits. Our results highlight the importance of revisiting the analysis of known exoplanetary systems when new data become available, particularly given the possibility that poorly-sampled data might previously have led to the detection of a 'false-positive' single eccentric planet, when the system in question actually contains two (or more) planets on near-circular orbits.

astro-ph.EP↗

arXiv2017

The Pan-Pacific Planet Search VII: The most eccentric planet orbiting a giant star

Radial velocity observations from three instruments reveal the presence of a 4 M_jup planet candidate orbiting the K giant HD 76920. HD 76920b has an orbital eccentricity of 0.856$\pm$0.009, making it the most eccentric planet known to orbit an evolved star. There is no indication that HD 76920 has an unseen binary companion, suggesting a scattering event rather than Kozai oscillations as a probable culprit for the observed eccentricity. The candidate planet currently approaches to about four stellar radii from its host star, and is predicted to be engulfed on a $\sim$100 Myr timescale due to the combined effects of stellar evolution and tidal interactions.

astro-ph.EP↗

arXiv2016

The Pan-Pacific Planet Search VI: Giant planets orbiting HD 86950 and HD 222076

We report the detection of two new planets orbiting the K giants HD 86950 and HD 222076, based on precise radial velocities obtained with three instruments: AAT/UCLES, FEROS, and CHIRON. HD 86950b has a period of 1270$\pm$57 days at $a=2.72\pm$0.08 AU, and m sin $i=3.6\pm$0.7 Mjup. HD 222076b has $P=871\pm$19 days at $a=1.83\pm$0.03 AU, and m sin $i=1.56\pm$0.11 Mjup. These two giant planets are typical of the population of planets known to orbit evolved stars. In addition, we find a high-amplitude periodic velocity signal ($K\sim$50 m/s) in HD 29399, and show that it is due to stellar variability rather than Keplerian reflex motion. We also investigate the relation between planet occurrence and host-star metallicity for the 164-star Pan-Pacific Planet Search sample of evolved stars. In spite of the small sample of PPPS detections, we confirm the trend of increasing planet occurrence as a function of metallicity found by other studies of planets orbiting evolved stars.

astro-ph.EP↗

arXiv2015

Three planets orbiting Wolf 1061

We use archival HARPS spectra to detect three planets orbiting the M3 dwarf Wolf1061 (GJ 628). We detect a 1.36 Mearth minimum-mass planet with an orbital period P = 4.888d (Wolf1061b), a 4.25 Mearth minimum-mass planet with orbital period P = 17.867d (Wolf1061c), and a likely 5.21 Mearth minimum-mass planet with orbital period P = 67.274d (Wolf1061d). All of the planets are of sufficiently low mass that they may be rocky in nature. The 17.867d planet falls within the habitable zone for Wolf 1061 and the 67.274d planet falls just outside the outer boundary of the habitable zone. There are no signs of activity observed in the bisector spans, cross-correlation full-width-half-maxima, Calcium H & K indices, NaD indices, or H-alpha indices near the planetary periods. We use custom methods to generate a cross-correlation template tailored to the star. The resulting velocities do not suffer the strong annual variation observed in the HARPS DRS velocities. This differential technique should deliver better exploitation of the archival HARPS data for the detection of planets at extremely low amplitudes.

astro-ph.EP↗