arXiv subjects

Ting Xiao

Publications and source records attributed to Ting Xiao.

At least 37 records · Page 2Linked to original sources

arXiv2025

Learn as Individuals, Evolve as a Team: Multi-agent LLMs Adaptation in Embodied Environments

Large language models (LLMs) possess extensive knowledge bases and strong reasoning capabilities, making them promising tools for complex, multi-agent planning in embodied environments. However, despite LLMs' advanced abilities and the sophisticated modular design of agentic methods, existing LLM-based planning algorithms remain limited by weak adaptation capabilities to multi-agent embodied scenarios. We address this limitation by introducing a framework that enables LLM agents to learn and evolve both before and during test time, equipping them with environment-relevant knowledge for better planning and enhanced communication for improved cooperation. Inspired by centralized training with decentralized execution in multi-agent reinforcement learning, we propose a \textit{Learn as Individuals, Evolve as a Team (LIET)} paradigm for multi-agent LLMs adaptation. At the individual level, LLM agents learn a local utility function from exploratory datasets to better comprehend the embodied environment, which is then queried during test time to support informed decision-making. At the team level, LLM agents collaboratively and iteratively maintain and update a shared cooperation knowledge list based on new experiences, using it to guide more effective communication. By combining individual learning with team evolution, LIET enables comprehensive and flexible adaptation for LLM agents. Our experiments on Communicative Watch-And-Help and ThreeD-World Multi-Agent Transport benchmarks demonstrate that LIET, instantiated with both LLaMA and GPT-4o, outperforms existing baselines and exhibits strong cooperative planning abilities.

cs.MA↗

arXiv2025

Online Iterative Self-Alignment for Radiology Report Generation

Radiology Report Generation (RRG) is an important research topic for relieving radiologist' heavy workload. Existing RRG models mainly rely on supervised fine-tuning (SFT) based on different model architectures using data pairs of radiological images and corresponding radiologist-annotated reports. Recent research has shifted focus to post-training improvements, aligning RRG model outputs with human preferences using reinforcement learning (RL). However, the limited data coverage of high-quality annotated data poses risks of overfitting and generalization. This paper proposes a novel Online Iterative Self-Alignment (OISA) method for RRG that consists of four stages: self-generation of diverse data, self-evaluation for multi-objective preference data,self-alignment for multi-objective optimization and self-iteration for further improvement. Our approach allows for generating varied reports tailored to specific clinical objectives, enhancing the overall performance of the RRG model iteratively. Unlike existing methods, our frame-work significantly increases data quality and optimizes performance through iterative multi-objective optimization. Experimental results demonstrate that our method surpasses previous approaches, achieving state-of-the-art performance across multiple evaluation metrics.

cs.CV↗

arXiv2025

A Multimodal Multi-Agent Framework for Radiology Report Generation

Radiology report generation (RRG) aims to automatically produce diagnostic reports from medical images, with the potential to enhance clinical workflows and reduce radiologists' workload. While recent approaches leveraging multimodal large language models (MLLMs) and retrieval-augmented generation (RAG) have achieved strong results, they continue to face challenges such as factual inconsistency, hallucination, and cross-modal misalignment. We propose a multimodal multi-agent framework for RRG that aligns with the stepwise clinical reasoning workflow, where task-specific agents handle retrieval, draft generation, visual analysis, refinement, and synthesis. Experimental results demonstrate that our approach outperforms a strong baseline in both automatic metrics and LLM-based evaluations, producing more accurate, structured, and interpretable reports. This work highlights the potential of clinically aligned multi-agent frameworks to support explainable and trustworthy clinical AI applications.

cs.AI↗

arXiv2024

Radiology Report Generation via Multi-objective Preference Optimization

Automatic Radiology Report Generation (RRG) is an important topic for alleviating the substantial workload of radiologists. Existing RRG approaches rely on supervised regression based on different architectures or additional knowledge injection,while the generated report may not align optimally with radiologists' preferences. Especially, since the preferences of radiologists are inherently heterogeneous and multidimensional, e.g., some may prioritize report fluency, while others emphasize clinical accuracy. To address this problem,we propose a new RRG method via Multi-objective Preference Optimization (MPO) to align the pre-trained RRG model with multiple human preferences, which can be formulated by multi-dimensional reward functions and optimized by multi-objective reinforcement learning (RL). Specifically, we use a preference vector to represent the weight of preferences and use it as a condition for the RRG model. Then, a linearly weighed reward is obtained via a dot product between the preference vector and multi-dimensional reward. Next,the RRG model is optimized to align with the preference vector by optimizing such a reward via RL. In the training stage,we randomly sample diverse preference vectors from the preference space and align the model by optimizing the weighted multi-objective rewards, which leads to an optimal policy on the entire preference space. When inference,our model can generate reports aligned with specific preferences without further fine-tuning. Extensive experiments on two public datasets show the proposed method can generate reports that cater to different preferences in a single model and achieve state-of-the-art performance.

cs.LG↗

arXiv2024

Constrained Ensemble Exploration for Unsupervised Skill Discovery

Unsupervised Reinforcement Learning (RL) provides a promising paradigm for learning useful behaviors via reward-free per-training. Existing methods for unsupervised RL mainly conduct empowerment-driven skill discovery or entropy-based exploration. However, empowerment often leads to static skills, and pure exploration only maximizes the state coverage rather than learning useful behaviors. In this paper, we propose a novel unsupervised RL framework via an ensemble of skills, where each skill performs partition exploration based on the state prototypes. Thus, each skill can explore the clustered area locally, and the ensemble skills maximize the overall state coverage. We adopt state-distribution constraints for the skill occupancy and the desired cluster for learning distinguishable skills. Theoretical analysis is provided for the state entropy and the resulting skill distributions. Based on extensive experiments on several challenging tasks, we find our method learns well-explored ensemble skills and achieves superior performance in various downstream tasks compared to previous methods.

cs.LG↗

arXiv2024

The ALMaQUEST Survey XV: The Dependence of the Molecular-to-Atomic Gas Ratios on Resolved Optical Diagnostics

The atomic-to-molecular gas conversion is a critical step in the baryon cycle of galaxies, which sets the initial conditions for subsequent star formation and influences the multi-phase interstellar medium. We compiled a sample of 94 nearby galaxies with observations of multi-phase gas contents by utilizing public H I, CO, and optical IFU data from the MaNGA survey together with new FAST H I observations. In agreement with previous results, our sample shows that the global molecular-to-atomic gas ratio ($R_{\rm mol} \equiv$ log $M_{\rm H_2}/M_{\rm H\ I}$) is correlated with the global stellar mass surface density $μ_*$ with a Kendall's $τ$ coefficient of 0.25 and $p < 10^{-3}$, less tightly but still correlated with stellar mass and NUV$-$ r color, and not related to the specific star formation rate (sSFR). The cold gas distribution and kinematics inferred from the H I and CO global profile asymmetry and shape do not significantly rely on $R_{\rm mol}$. Thanks to the availability of kpc-scale observations of MaNGA, we decompose galaxies into H II, composite, and AGN-dominated regions by using the BPT diagrams. With increasing $R_{\rm mol}$, the fraction of H II regions within 1.5 effective radius decreases slightly; the density distribution in the spatially resolved BPT diagram also changes significantly, suggesting changes in metallicity and ionization states. Galaxies with high $R_{\rm mol}$ tend to have high oxygen abundance, both at one effective radius with a Kendall's $τ$ coefficient of 0.37 ($p < 10^{-3}$) and their central regions. Among all parameters investigated here, the oxygen abundance at one effective radius has the strongest relation with global $R_{\rm mol}$, but the dependence of gas conversion on gas distribution and galaxy ionization states is weak.

astro-ph.GA↗

arXiv2024

On the existence, rareness and uniqueness of quenched HI-rich galaxies in the local Universe

Using data from ALFALFA, xGASS, HI-MaNGA and the Sloan Digital Sky Survey (SDSS), we identify a sample of 47 "red but HI-rich"(RR) galaxies with $NUV-r > 5$ and unusually high HI-to-stellar mass ratios. We compare the optical properties and local environments between the RR galaxies and a control sample of "red and HI-normal"(RN) galaxies that are matched in stellar mass and color. The two samples are similar in the optical properties typical of massive red (quenched) galaxies in the local Universe. The RR sample tends to be associated with slightly lower-density environments and has lower clustering amplitudes and smaller neighbor counts at scales from several kiloparsecs to a few Megaparsecs. The results are consistent with the RR galaxies preferentially being located at the center of low-mass halos, with a median halo mass $\sim 10^{12}h^{-1}M_{\odot}$ compared to $\sim 10^{12.5}h^{-1}M_{\odot}$ for the RN sample. This result is confirmed by the SDSS group catalog which reveals a central fraction of 89% for the RR sample, compared to $\sim 60\%$ for the RN sample. If assumed to follow the HI size-mass relation of normal galaxies, the RR galaxies have an average HI-to-optical radius ratio of $R_{HI}/R_{90}\sim 4$, four times the average ratio for the RN sample. We compare our RR sample with similar samples in previous studies, and quantify the population of RR galaxies using the SDSS complete sample. We conclude that the RR galaxies form a unique but rare population, accounting for only a small fraction of the massive quiescent galaxy population. We discuss the formation scenarios of the RR galaxies.

astro-ph.GA↗

arXiv2023

Improving Few-shot Image Generation by Structural Discrimination and Textural Modulation

Few-shot image generation, which aims to produce plausible and diverse images for one category given a few images from this category, has drawn extensive attention. Existing approaches either globally interpolate different images or fuse local representations with pre-defined coefficients. However, such an intuitive combination of images/features only exploits the most relevant information for generation, leading to poor diversity and coarse-grained semantic fusion. To remedy this, this paper proposes a novel textural modulation (TexMod) mechanism to inject external semantic signals into internal local representations. Parameterized by the feedback from the discriminator, our TexMod enables more fined-grained semantic injection while maintaining the synthesis fidelity. Moreover, a global structural discriminator (StructD) is developed to explicitly guide the model to generate images with reasonable layout and outline. Furthermore, the frequency awareness of the model is reinforced by encouraging the model to distinguish frequency signals. Together with these techniques, we build a novel and effective model for few-shot image generation. The effectiveness of our model is identified by extensive experiments on three popular datasets and various settings. Besides achieving state-of-the-art synthesis performance on these datasets, our proposed techniques could be seamlessly integrated into existing models for a further performance boost.

cs.CV↗

arXiv2022

Conditional HI mass functions and the HI-to-halo mass relation in the local Universe

We present a new HI mass estimator which relates the HI-to-stellar mass ratio to four galaxy properties: stellar surface mass density, color index $u-r$, stellar mass and concentration index, with the scatter of individual galaxies around the mean HI mass modeled with a Gaussian distribution. We calibrate the estimator using the xGASS sample, including both HI detection and non-detection, and constrain the model parameters through Bayesian inferences. Tests with mock catalogs demonstrate that our estimator provides unbiased HI masses for optical samples like the SDSS, thus suitable for statistical studies of HI gas contents in galaxies and dark matter halos. We apply our estimator to the SDSS spectroscopic sample to estimate the local HI mass function (HIMF), the conditional HI mass function (CHIMF) in galaxy groups and the HI-halo mass (HIHM) relation. Our HIMF agrees with the ALFALFA measurements at $M_{HI}\gtrsim 5\times 10^9M_{\odot}$, but with higher amplitude and a steeper slope at lower masses. We show that this discrepancy is caused primarily by the cosmic variance which is corrected for the SDSS sample but not for the ALFALFA. The CHIMFs for all halo masses can be described by a single Schechter function, and this is true for red, blue and satellite galaxies. For central galaxies the CHIMFs show a double-Gaussian profile, with the two components contributed by the red and blue galaxies, respectively. The total HI mass in a group increases monotonically with halo mass. The HI mass of central galaxies in galaxy groups increases rapidly with halo mass only at $M_h\lesssim10^{12}M_{\odot}$, while the mass dependence becomes much weaker at higher halo masses. The observed HI-halo mass relation is not reproduced by current hydrodynamic simulations and semi-analytic models of galaxy formation.

astro-ph.GA↗

arXiv2022

The cold gas and dust properties of red star-forming galaxies

We study the cold gas and dust properties for a sample of red star forming galaxies called "red misfits." We collect single-dish CO observations and HI observations from representative samples of low-redshift galaxies, as well as our own JCMT CO observations of red misfits. We also obtain SCUBA-2 850 um observations for a subset of these galaxies. With these data we compare the molecular gas, total cold gas, and dust properties of red misfits against those of their blue counterparts ("blue actives") taking non-detections into account using a survival analysis technique. We compare these properties at fixed position in the log SFR-log M* plane, as well as versus offset from the star-forming main sequence. Compared to blue actives, red misfits have slightly longer molecular gas depletion times, similar total gas depletion times, significantly lower molecular- and total-gas mass fractions, lower dust-to-stellar mass ratios, similar dust-to-gas ratios, and a significantly flatter slope in the $\log M_\mathrm{mol}$-$\log M_\star$ plane. Our results suggest that red misfits as a population are likely quenching due to a shortage in gas supply.

astro-ph.GA↗

arXiv2022

Stock2Vec: An Embedding to Improve Predictive Models for Companies

Building predictive models for companies often relies on inference using historical data of companies in the same industry sector. However, companies are similar across a variety of dimensions that should be leveraged in relevant prediction problems. This is particularly true for large, complex organizations which may not be well defined by a single industry and have no clear peers. To enable prediction using company information across a variety of dimensions, we create an embedding of company stocks, Stock2Vec, which can be easily added to any prediction model that applies to companies with associated stock prices. We describe the process of creating this rich vector representation from stock price fluctuations, and characterize what the dimensions represent. We then conduct comprehensive experiments to evaluate this embedding in applied machine learning problems in various business contexts. Our experiment results demonstrate that the four features in the Stock2Vec embedding can readily augment existing cross-company models and enhance cross-company predictions.

cs.LG↗

arXiv2021

Importance Weighted Adversarial Discriminative Transfer for Anomaly Detection

Previous transfer methods for anomaly detection generally assume the availability of labeled data in source or target domains. However, such an assumption is not valid in most real applications where large-scale labeled data are too expensive. Therefore, this paper proposes an importance weighted adversarial autoencoder-based method to transfer anomaly detection knowledge in an unsupervised manner, particularly for a rarely studied scenario where a target domain has no labeled normal/abnormal data while only normal data from a related source domain exist. Specifically, the method learns to align the distributions of normal data in both source and target domains, but leave the distribution of abnormal data in the target domain unchanged. In this way, an obvious gap can be produced between the distributions of normal and abnormal data in the target domain, therefore enabling the anomaly detection in the domain. Extensive experiments on multiple synthetic datasets and the UCSD benchmark demonstrate the effectiveness of our approach.

cs.LG↗

arXiv2020

ALMaQUEST -- IV. The ALMA-MaNGA QUEnching and STar formation (ALMaQUEST) Survey

The ALMaQUEST (ALMA-MaNGA QUEnching and STar formation) survey is a program with spatially-resolved $^{12}$CO(1-0) measurements obtained with the Atacama Large Millimeter Array (ALMA) for 46 galaxies selected from the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) DR15 optical integral-field spectroscopic survey. The aim of the ALMaQUEST survey is to investigate the dependence of star formation activity on the cold molecular gas content at kpc scales in nearby galaxies. The sample consists of galaxies spanning a wide range in specific star formation rate (sSFR), including starburst (SB), main-sequence (MS), and green valley (GV) galaxies. In this paper, we present the sample selection and characteristics of the ALMA observations, and showcase some of the key results enabled by the combination of spatially-matched stellar populations and gas measurements. Considering the global (aperture-matched) stellar mass, molecular gas mass, and star formation rate of the sample, we find that the sSFR depends on both the star formation efficiency (SFE) and the molecular gas fraction ($f_{\rm H_{2}}$), although the correlation with the latter is slightly weaker. Furthermore, the dependence of sSFR on the molecular gas content (SFE or $f_{\rm H_{2}}$) is stronger than that on either the atomic gas fraction or the molecular-to-atomic gas fraction, albeit with the small HI sample size. On kpc scales, the variations in both SFE and $f_{\rm H_{2}}$ within individual galaxies can be as large as 1-2 dex thereby demonstrating that the availability of spatially-resolved observations is essential to understand the details of both star formation and quenching processes.

astro-ph.GA↗

arXiv2020

SDSS-IV MaNGA: the indispensable role of bars in enhancing the central star formation of low-$z$ galaxies

We analyse two-dimensional maps and radial profiles of EW(H$α$), EW(H$δ_A$), and D$_n$(4000) of low-redshift galaxies using integral field spectroscopy from the MaNGA survey. Out of $\approx1400$ nearly face-on late-type galaxies with a redshift $z<0.05$, we identify 121 "turnover" galaxies that each have a central upturn in EW(H$α$), EW(H$δ_A$) and/or a central drop in D$_n$(4000), indicative of ongoing/recent star formation. The turnover features are found mostly in galaxies with a stellar mass above $\sim$10$^{10}$ M$_{\odot}$ and NUV-$r$ colour less than $\approx5$. The majority of the turnover galaxies are barred, with a bar fraction of 89$\pm$3\%. Furthermore, for barred galaxies the radius of the central turnover region is found to tightly correlate with one third of the bar length. Comparing the observed and the inward extrapolated star formation rate surface density, we estimate that the central SFR have been enhanced by an order of magnitude. Conversely, only half of the barred galaxies in our sample have a central turnover feature, implying that the presence of a bar is not sufficient to lead to a central SF enhancement. We further examined the SF enhancement in paired galaxies, as well as the local environment, finding no relation. This implies that environment is not a driving factor for central SF enhancement in our sample. Our results reinforce both previous findings and theoretical expectation that galactic bars play a crucial role in the secular evolution of galaxies by driving gas inflow and enhancing the star formation and bulge growth in the center.

astro-ph.GA↗

arXiv2019

Estimating the molecular gas mass of low-redshift galaxies from a combination of mid-infrared luminosity and optical properties

We present CO(J=1-0) and/or CO(J=2-1) spectroscopy for 31 galaxies selected from the ongoing MaNGA survey, obtained with multiple telescopes. This sample is combined with CO observations from the literature to study the correlation of the CO luminosities ($L_{\rm CO(1-0)}$) with the mid-infrared luminosities at 12 ($L_{12 μm}$) and 22 $μ$m ($L_{\rm 22 μm}$), as well as the dependence of the residuals on a variety of galaxy properties. The correlation with $L_{\rm 12 μm}$ is tighter and more linear, but galaxies with relatively low stellar masses and blue colors fall significantly below the mean $L_{\rm CO(1-0)}-L_{\rm 12μm}$ relation. We propose a new estimator of the CO(1-0) luminosity (and thus the total molecular gas mass) that is a linear combination of three parameters: $L_{\rm 12 μm}$, $M_\ast$ and $g-r$. We show that, with a scatter of only 0.18 dex in log $(L_{\rm CO(1-0)})$, this estimator provides unbiased estimates for galaxies of different properties and types. An immediate application of this estimator to a compiled sample of galaxies with only CO(J=2-1) observations yields a distribution of the CO(J=2-1) to CO(J=1-0) luminosity ratios ($R21$) that agrees well with the distribution of real observations, in terms of both the median and the shape. Application of our estimator to the current MaNGA sample reveals a gas-poor population of galaxies that are predominantly early-type and show no correlation between molecular gas-to-stellar mass ratio and star formation rate, in contrast to gas-rich galaxies. We also provide alternative estimators with similar scatters, based on $r$ and/or $z$ band luminosities instead of $M_\ast$. These estimators serve as cheap and convenient $M_{\rm mol}$ proxies to be potentially applied to large samples of galaxies, thus allowing statistical studies of gas-related processes of galaxies.

astro-ph.GA↗

arXiv2019

JINGLE V: Dust properties of nearby galaxies derived from hierarchical Bayesian SED fitting

We study the dust properties of 192 nearby galaxies from the JINGLE survey using photometric data in the 22-850micron range. We derive the total dust mass, temperature T and emissivity index beta of the galaxies through the fitting of their spectral energy distribution (SED) using a single modified black-body model (SMBB). We apply a hierarchical Bayesian approach that reduces the known degeneracy between T and beta. Applying the hierarchical approach, the strength of the T-beta anti-correlation is reduced from a Pearson correlation coefficient R=-0.79 to R=-0.52. For the JINGLE galaxies we measure dust temperatures in the range 17-30 K and dust emissivity indices beta in the range 0.6-2.2. We compare the SMBB model with the broken emissivity modified black-body (BMBB) and the two modified black-bodies (TMBB) models. The results derived with the SMBB and TMBB are in good agreement, thus applying the SMBB, which comes with fewer free parameters, does not penalize the measurement of the cold dust properties in the JINGLE sample. We investigate the relation between T and beta and other global galaxy properties in the JINGLE and Herschel Reference Survey (HRS) sample. We find that beta correlates with the stellar mass surface density (R=0.62) and anti-correlates with the HI mass fraction (M(HI)/M*, R=-0.65), whereas the dust temperature correlates strongly with the SFR normalized by the dust mass (R=0.73). These relations can be used to estimate T and beta in galaxies with insufficient photometric data available to measure them directly through SED fitting.

astro-ph.GA↗

arXiv2019

SDSS-IV MaNGA: Spatial Evolution of Star Formation Triggered by Galaxy Interactions

Galaxy interaction is considered a key driver of galaxy evolution and star formation (SF) history. In this paper, we present an empirical picture of the radial extent of interaction-triggered SF along the merger sequence. The samples under study are drawn from the integral field spectroscopy (IFS) survey SDSS-IV MaNGA, including 205 star-forming galaxies in pairs/mergers and ~1350 control galaxies. For each galaxy in pairs, the merger stage is identified according to its morphological signatures: incoming phase, at first pericenter passage, at apocenter, in merging phase, and in final coalescence. The effect of interactions is quantified by the global and spatially resolved SF rate relative to the SF rate of a control sample selected for each individual galaxy ($Δ$logSFR and $Δ$logsSFR(r), respectively). Analysis of the radial $Δ$logsSFR(r) distributions shows that galaxy interactions have no significant impact on the $Δ$logsSFR(r) during the incoming phase. Right after the first pericenter passage, the radial $Δ$logsSFR(r) profile decreases steeply from enhanced to suppressed activity for increasing galactocentric radius. Later on, SF is enhanced on a broad spatial scale out to the maximum radius we explore (~6.7 kpc) and the enhancement is in general centrally peaked. The extended SF enhancement is also observed for systems at their apocenters and in the coalescence phase, suggesting that interaction-triggered SF is not restricted to the central region of a galaxy. Further explorations of a wide range in parameter space of merger configurations (e.g., mass ratio) are required to constrain the whole picture of interaction-triggered SF.

astro-ph.GA↗

arXiv2019

JINGLE, a JCMT legacy survey of dust and gas for galaxy evolution studies: II. SCUBA-2 850 μm data reduction and dust flux density catalogues

We present the SCUBA-2 850 $μm$ component of JINGLE, the new JCMT large survey for dust and gas in nearby galaxies, which with 193 galaxies is the largest targeted survey of nearby galaxies at 850 $μm$. We provide details of our SCUBA-2 data reduction pipeline, optimised for slightly extended sources, and including a calibration model adjusted to match conventions used in other far-infrared data. We measure total integrated fluxes for the entire JINGLE sample in 10 infrared/submillimetre bands, including all WISE, Herschel-PACS, Herschel-SPIRE and SCUBA-2 850 $μm$ maps, statistically accounting for the contamination by CO(J=3-2) in the 850 $μm$ band. Of our initial sample of 193 galaxies, 191 are detected at 250 $μm$ with a $\geq$ 5$σ$ significance. In the SCUBA-2 850 $μm$ band we detect 126 galaxies with $\geq$ 3$σ$ significance. The distribution of the JINGLE galaxies in far-infrared/sub-millimetre colour-colour plots reveals that the sample is not well fit by single modified-blackbody models that assume a single dust-emissivity index $(β)$. Instead, our new 850 $μm$ data suggest either that a large fraction of our objects require $β< 1.5$, or that a model allowing for an excess of sub-mm emission (e.g., a broken dust emissivity law, or a very cold dust component 10 K) is required. We provide relations to convert far-infrared colours to dust temperature and $β$ for JINGLE-like galaxies. For JINGLE the FIR colours correlate more strongly with star-formation rate surface-density rather than the stellar surface-density, suggesting heating of dust is greater due to younger rather than older stellar-populations, consistent with the low proportion of early-type galaxies in the sample.

astro-ph.GA↗