arXiv subjects

Yunyi Shen

Publications and source records attributed to Yunyi Shen.

At least 19 recordsLinked to original sources

arXiv2026

Radiative Transfer Modeling of Stripped-envelope Supernovae II: Neural Network Emulation of Light Curves

We present the first neural-network emulator of stripped-envelope supernova (SESN) lightcurves, trained on a grid of 4499 light curves simulated with the radiative transfer (RT) code sedona. Using this emulator, we show that m_ni, m_ejecta, the ejecta velocity profile, and the degree of Ni-56 mixing can all be inferred from multiband lightcurves. We find that the degeneracy between ejecta mass and ejecta velocity is substantially weaker with this emulator than in traditional semianalytical models. The emulator is able to independently constrain the influence of ejecta mass and of ejecta velocity on the resulting lightcurve, rather than making them degenerate by design as traditional semianalytical models do. We additionally show that this inference is significantly more accurate than that done by the classical Arnett model for both simulated ZTF-like and LSST-like lightcurves. Finally, we present lightcurves fits to three well-studied SESNe: SN~1994I, SN~2007gr, and iPTF13bvn, constraining their m_ni, m_ejecta, and Ni-56 mixing.

astro-ph.HE↗

arXiv2025

Minuet: A Diffusion Autoencoder for Compact Semantic Compression of Multi-Band Galaxy Images

The Vera C. Rubin Observatory is slated to observe nearly 20 billion galaxies during its decade-long Legacy Survey of Space and Time. The rich imaging data it collects will be an invaluable resource for probing galaxy evolution across cosmic time, characterizing the host galaxies of transient phenomena, and identifying novel populations of anomalous systems. While machine learning models have shown promise for extracting galaxy features from multi-band astronomical imaging, the large dimensionality of the learned latent space presents a challenge for mechanistic interpretability studies. In this work, we present Minuet, a low-dimensional diffusion autoencoder for multi-band galaxy imaging. Minuet is trained to reconstruct 72x72-pixel $grz$ image cutouts of 6M galaxies within $z<1$ from the Dark Energy Camera Legacy Survey using only five latent dimensions. By using a diffusion model conditioned on the transformer-based autoencoder's output for image reconstruction, we achieve semantically-meaningful latent representations of galaxy images while still allowing for high-fidelity, probabilistic reconstructions. We train a series of binary classifiers on Minuet's latent features to quantify their connection to morphological labels from Galaxy Zoo, and a conditional flow to produce posterior distributions of SED-derived redshifts, stellar masses, and star-formation rates. We further show the value of Minuet for nearest neighbor searches in the learned latent space. Minuet provides strong evidence for the low intrinsic dimensionality of galaxy imaging, and introduces a class of astrophysical models that produce highly compact representations for diverse science goals.

astro-ph.IM↗

arXiv2025

Explicit modeling of density dependence in spatial capture-recapture models

Density dependence occurs at the individual level and thus is greatly influenced by spatial local heterogeneity in habitat conditions. However, density dependence is often evaluated at the population level, leading to difficulties or even controversies in detecting such a process. Bayesian individual-based models such as spatial capture-recapture (SCR) models provide opportunities to study density dependence at the individual level, but such an approach remains to be developed and evaluated. In this study, we developed a SCR model that links habitat use to apparent survival and recruitment through density dependent processes at the individual level. Using simulations, we found that the model can properly inform habitat use, but tends to underestimate the effect of density dependence on apparent survival and recruitment. The reason for such underestimations is likely due to the difficulties of the current model in identifying the locations of unobserved individuals without using environmental covariates to inform these locations. How to accurately estimate the locations of unobserved individuals, and thus density dependence, remains a challenging topic in spatial statistics and statistical ecology.

q-bio.PE↗

arXiv2025

Diffusion Autoencoders with Perceivers for Long, Irregular and Multimodal Astronomical Sequences

Self-supervised learning has become a central strategy for representation learning, but the majority of architectures used for encoding data have only been validated on regularly-sampled inputs such as images, audios. and videos. In many scientific domains, data instead arrive as long, irregular, and multimodal sequences. To extract semantic information from these data, we introduce the Diffusion Autoencoder with Perceivers (daep). daep tokenizes heterogeneous measurements, compresses them with a Perceiver encoder, and reconstructs them with a Perceiver-IO diffusion decoder, enabling scalable learning in diverse data settings. To benchmark the daep architecture, we adapt the masked autoencoder to a Perceiver encoder/decoder design, and establish a strong baseline (maep) in the same architectural family as daep. Across diverse spectroscopic and photometric astronomical datasets, daep achieves lower reconstruction errors, produces more discriminative latent spaces, and better preserves fine-scale structure than both VAE and maep baselines. These results establish daep as an effective framework for scientific domains where data arrives as irregular, heterogeneous sequences.

stat.ML↗

arXiv2025

Dropping Just a Handful of Preferences Can Change Top Large Language Model Rankings

We propose a method for evaluating the robustness of widely used LLM ranking systems -- variants of a Bradley--Terry model -- to dropping a worst-case very small fraction of preference data. Our approach is computationally fast and easy to adopt. When we apply our method to matchups from popular LLM ranking platforms, including Chatbot Arena and derivatives, we find that the rankings of top-performing models can be remarkably sensitive to the removal of a small fraction of preferences; for instance, dropping just 0.003% of human preferences can change the top-ranked model on Chatbot Arena. Our robustness check identifies the specific preferences most responsible for such ranking flips, allowing for inspection of these influential preferences. We observe that the rankings derived from MT-bench preferences are notably more robust than those from Chatbot Arena, likely due to MT-bench's use of expert annotators and carefully constructed prompts. Finally, we find that neither rankings based on crowdsourced human evaluations nor those based on LLM-as-a-judge preferences are systematically more sensitive than the other.

stat.ML↗

arXiv2025

Mixture-of-Expert Variational Autoencoders for Cross-Modality Embedding of Type Ia Supernova Data

Time-domain astrophysics relies on heterogeneous and multi-modal data. Specialized models are often constructed to extract information from a single modality, but this approach ignores the wealth of cross-modality information that may be relevant for the tasks to which the model is applied. In this work, we propose a multi-modal, mixture-of-expert variational autoencoder to learn a joint embedding for supernova light curves and spectra. Our method, which is inspired by the Perceiver architecture, natively accommodates variable-length inputs and the irregular temporal sampling inherent to supernova light curves. We train our model on radiative transfer simulations and validate its performance on cross-modality reconstruction of supernova spectra and physical parameters from the simulation. Our model achieves superior performance in cross-modality generation to nearest-neighbor searches in a contrastively-trained latent space, showing its promise for constructing informative latent representations of multi-modal astronomical datasets.

astro-ph.IM↗

arXiv2025

Approximations to worst-case data dropping: unmasking failure modes

A data analyst might worry about generalization if dropping a very small fraction of data points from a study could change its substantive conclusions. Checking this non-robustness directly poses a combinatorial optimization problem and is intractable even for simple models and moderate data sizes. Recently various authors have proposed a diverse set of approximations to detect this non-robustness. In the present work, we show that, even in a setting as simple as ordinary least squares (OLS) linear regression, many of these approximations can fail to detect (true) non-robustness in realistic data arrangements. We focus on OLS in the present work due its widespread use and since some approximations work only for OLS. Across our synthetic and real-world data sets, we find that a simple recursive greedy algorithm is the sole algorithm that does not fail any of our tests and also that it can be orders of magnitude faster to run than some competitors.

stat.ME↗

arXiv2025

OpenReview Should be Protected and Leveraged as a Community Asset for Research in the Era of Large Language Models

In the era of large language models (LLMs), high-quality, domain-rich, and continuously evolving datasets capturing expert-level knowledge, core human values, and reasoning are increasingly valuable. This position paper argues that OpenReview -- the continually evolving repository of research papers, peer reviews, author rebuttals, meta-reviews, and decision outcomes -- should be leveraged more broadly as a core community asset for advancing research in the era of LLMs. We highlight three promising areas in which OpenReview can uniquely contribute: enhancing the quality, scalability, and accountability of peer review processes; enabling meaningful, open-ended benchmarks rooted in genuine expert deliberation; and supporting alignment research through real-world interactions reflecting expert assessment, intentions, and scientific values. To better realize these opportunities, we suggest the community collaboratively explore standardized benchmarks and usage guidelines around OpenReview, inviting broader dialogue on responsible data use, ethical considerations, and collective stewardship.

cs.CY↗

arXiv2025

Oh SnapMMD! Forecasting Stochastic Dynamics Beyond the Schrödinger Bridge's End

Scientists often want to make predictions beyond the observed time horizon of "snapshot" data following latent stochastic dynamics. For example, in time course single-cell mRNA profiling, scientists have access to cellular transcriptional state measurements (snapshots) from different biological replicates at different time points, but they cannot access the trajectory of any one cell because measurement destroys the cell. Researchers want to forecast (e.g.) differentiation outcomes from early state measurements of stem cells. Recent Schrödinger-bridge (SB) methods are natural for interpolating between snapshots. But past SB papers have not addressed forecasting -- likely since existing methods either (1) reduce to following pre-set reference dynamics (chosen before seeing data) or (2) require the user to choose a fixed, state-independent volatility since they minimize a Kullback-Leibler divergence. Either case can lead to poor forecasting quality. In the present work, we propose a new framework, SnapMMD, that learns dynamics by directly fitting the joint distribution of both state measurements and observation time with a maximum mean discrepancy (MMD) loss. Unlike past work, our method allows us to infer unknown and state-dependent volatilities from the observed data. We show in a variety of real and synthetic experiments that our method delivers accurate forecasts. Moreover, our approach allows us to learn in the presence of incomplete state measurements and yields an $R^2$-style statistic that diagnoses fit. We also find that our method's performance at interpolation (and general velocity-field reconstruction) is at least as good as (and often better than) state-of-the-art in almost all of our experiments.

stat.ML↗

arXiv2025

Variational diffusion transformers for conditional sampling of supernovae spectra

Type Ia Supernovae (SNe Ia) have become the most precise distance indicators in astrophysics due to their incredible observational homogeneity. Increasing discovery rates, however, have revealed multiple sub-populations with spectroscopic properties that are both diverse and difficult to interpret using existing physical models. These peculiar events are hard to identify from sparsely sampled observations and can introduce systematics in cosmological analyses if not flagged early; they are also of broader importance for building a cohesive understanding of thermonuclear explosions. In this work, we introduce DiTSNe-Ia, a variational diffusion-based generative model conditioned on light curve observations and trained to reproduce the observed spectral diversity of SNe Ia. In experiments with realistic light curves and spectra from radiative transfer simulations, DiTSNe-Ia achieves significantly more accurate reconstructions than the widely used SALT3 templates across a broad range of observation phases (from 10 days before peak light to 30 days after it). DiTSNe-Ia yields a mean squared error of 0.108 across all phases-five times lower than SALT3's 0.508-and an after-peak error of just 0.0191, an order of magnitude smaller than SALT3's 0.305. Additionally, our model produces well-calibrated credible intervals with near-nominal coverage, particularly at post-peak phases. DiTSNe-Ia is a powerful tool for rapidly inferring the spectral properties of SNe Ia and other transient astrophysical phenomena for which a physical description does not yet exist.

astro-ph.IM↗

arXiv2025

Multi-marginal Schrödinger Bridges with Iterative Reference Refinement

Practitioners often aim to infer an unobserved population trajectory using sample snapshots at multiple time points. E.g., given single-cell sequencing data, scientists would like to learn how gene expression changes over a cell's life cycle. But sequencing any cell destroys that cell. So we can access data for any particular cell only at a single time point, but we have data across many cells. The deep learning community has recently explored using Schrödinger bridges (SBs) and their extensions in similar settings. However, existing methods either (1) interpolate between just two time points or (2) require a single fixed reference dynamic (often set to Brownian motion within SBs). But learning piecewise from adjacent time points can fail to capture long-term dependencies. And practitioners are typically able to specify a model family for the reference dynamic but not the exact values of the parameters within it. So we propose a new method that (1) learns the unobserved trajectories from sample snapshots across multiple time points and (2) requires specification only of a family of reference dynamics, not a single fixed one. We demonstrate the advantages of our method on simulated and real data.

stat.ML↗

arXiv2025

Consistent Validation for Predictive Methods in Spatial Settings

Spatial prediction tasks are key to weather forecasting, studying air pollution impacts, and other scientific endeavors. Determining how much to trust predictions made by statistical or physical methods is essential for the credibility of scientific conclusions. Unfortunately, classical approaches for validation fail to handle mismatch between locations available for validation and (test) locations where we want to make predictions. This mismatch is often not an instance of covariate shift (as commonly formalized) because the validation and test locations are fixed (e.g., on a grid or at select points) rather than i.i.d. from two distributions. In the present work, we formalize a check on validation methods: that they become arbitrarily accurate as validation data becomes arbitrarily dense. We show that classical and covariate-shift methods can fail this check. We propose a method that builds from existing ideas in the covariate-shift literature, but adapts them to the validation data at hand. We prove that our proposal passes our check. And we demonstrate its advantages empirically on simulated and real data.

stat.ML↗

arXiv2025

Wild posteriors in the wild

Bayesian posterior approximation has become more accessible to practitioners than ever, thanks to modern black-box software. While these tools provide highly accurate approximations with minimal user effort, certain posterior geometries remain notoriously difficult for standard methods. As a result, research into alternative approximation techniques continues to flourish. In many papers, authors validate their new approaches by testing them on posterior shapes deemed challenging or "wild." However, these shapes are not always directly linked to real-world applications where they naturally occur. In this note, we present examples of practical applications that give rise to some commonly used benchmark posterior shapes.

stat.CO↗

arXiv2025

Reviving The Classics: Active Reward Modeling in Large Language Model Alignment

Building neural reward models from human preferences is a pivotal component in reinforcement learning from human feedback (RLHF) and large language model alignment research. Given the scarcity and high cost of human annotation, how to select the most informative pairs to annotate is an essential yet challenging open problem. In this work, we highlight the insight that an ideal comparison dataset for reward modeling should balance exploration of the representation space and make informative comparisons between pairs with moderate reward differences. Technically, challenges arise in quantifying the two objectives and efficiently prioritizing the comparisons to be annotated. To address this, we propose the Fisher information-based selection strategies, adapt theories from the classical experimental design literature, and apply them to the final linear layer of the deep neural network-based reward modeling tasks. Empirically, our method demonstrates remarkable performance, high computational efficiency, and stability compared to other selection methods from deep learning and classical statistical literature across multiple open-source LLMs and datasets. Further ablation studies reveal that incorporating cross-prompt comparisons in active reward modeling significantly enhances labeling efficiency, shedding light on the potential for improved annotation strategies in RLHF.

cs.CL↗

arXiv2025

Reusing Embeddings: Reproducible Reward Model Research in Large Language Model Alignment without GPUs

Large Language Models (LLMs) have made substantial strides in structured tasks through Reinforcement Learning (RL), demonstrating proficiency in mathematical reasoning and code generation. However, applying RL in broader domains like chatbots and content generation -- through the process known as Reinforcement Learning from Human Feedback (RLHF) -- presents unique challenges. Reward models in RLHF are critical, acting as proxies that evaluate the alignment of LLM outputs with human intent. Despite advancements, the development of reward models is hindered by challenges such as computational heavy training, costly evaluation, and therefore poor reproducibility. We advocate for using embedding-based input in reward model research as an accelerated solution to those challenges. By leveraging embeddings for reward modeling, we can enhance reproducibility, reduce computational demands on hardware, improve training stability, and significantly reduce training and evaluation costs, hence facilitating fair and efficient comparisons in this active research area. We then show a case study of reproducing existing reward model ensemble research using embedding-based reward models. We discussed future avenues for research, aiming to contribute to safer and more effective LLM deployments.

cs.CL↗

arXiv2025

Rethinking Bradley-Terry Models in Preference-Based Reward Modeling: Foundations, Theory, and Alternatives

The Bradley-Terry (BT) model is a common and successful practice in reward modeling for Large Language Model (LLM) alignment. However, it remains unclear why this model -- originally developed for multi-player stochastic game matching -- can be adopted to convert pairwise response comparisons to reward values and make predictions. Especially given the fact that only a limited number of prompt-response pairs are sparsely compared with others. In this paper, we first revisit the foundations of using BT models in reward modeling, and establish the convergence rate of BT reward models based on deep neural networks using embeddings, providing a theoretical foundation for their use. Despite theoretically sound, we argue that the BT model is not a necessary choice from the perspective of downstream optimization. This is because a reward model only needs to preserve the correct ranking predictions through a monotonic transformation of the true reward. We highlight the critical concept of order consistency in reward modeling and demonstrate that the BT model possesses this property. Consequently, we propose a simple and straightforward upper-bound algorithm, compatible with off-the-shelf binary classifiers, as an alternative order-consistent reward modeling objective. To offer practical insights, we empirically evaluate the performance of these different reward modeling approaches across more than 12,000 experimental setups, using $6$ base LLMs, $2$ datasets, and diverse annotation designs that vary in quantity, quality, and pairing choices in preference annotations.

cs.AI↗

arXiv2024

Posterior contraction and uncertainty quantification for the multivariate spike-and-slab LASSO

We study the asymptotic properties of Deshpande et al.\ (2019)'s multivariate spike-and-slab LASSO (mSSL) procedure for simultaneous variable and covariance selection in the sparse multivariate linear regression problem. In that problem, $q$ correlated responses are regressed onto $p$ covariates and the mSSL works by placing separate spike-and-slab priors on the entries in the matrix of marginal covariate effects and off-diagonal elements in the upper triangle of the residual precision matrix. Under mild assumptions about these matrices, we establish the posterior contraction rate for the mSSL posterior in the asymptotic regime where both $p$ and $q$ diverge with $n.$ By ``de-biasing'' the corresponding MAP estimates, we obtain confidence intervals for each covariate effect and residual partial correlation. In extensive simulation studies, these intervals displayed close-to-nominal frequentist coverage in finite sample settings but tended to be substantially longer than those obtained using a version of the Bayesian bootstrap that randomly re-weights the prior. We further show that the de-biased intervals for individual covariate effects are asymptotically valid.

math.ST↗

arXiv2024

Detecting stellar flares in photometric data using hidden Markov models

We present a hidden Markov model (HMM) for discovering stellar flares in light curve data of stars. HMMs provide a framework to model time series data that are not stationary; they allow for systems to be in different states at different times and consider the probabilities that describe the switching dynamics between states. In the context of stellar flares discovery, we exploit the HMM framework by allowing the light curve of a star to be in one of three states at any given time step: Quiet, Firing, or Decaying. This three state HMM formulation is designed to enable straightforward identification of stellar flares, their duration, and associated uncertainty. This is crucial for estimating the flare's energy, and is useful for studies of stellar flare energy distributions. We combine our HMM with a celerite model that accounts for quasi periodic stellar oscillations. Through an injection recovery experiment, we demonstrate and evaluate the ability of our method to detect and characterize flares in stellar time series. We also show that the proposed HMM flags fainter and lower energy flares more easily than traditional sigma clipping methods. Lastly, we visually demonstrate that simultaneously conducting detrending and flare detection can mitigate biased estimations arising in multistage modelling approaches. Thus, this method paves a new way to calculating stellar flare energy. We conclude with an example application to one star observed by TESS, showing how the HMM compares with sigma clipping when using real data.

astro-ph.SR↗