arXiv subjects

Lin Xiao

Publications and source records attributed to Lin Xiao.

At least 19 recordsLinked to original sources

arXiv2026

The environmental dependence of mid-IR luminous dusty Supernovae

Using the Spitzer and WISE images, we discovered 42 mid-IR luminous dusty supernovae with local integral-field spectroscopy data. The observed mid-IR emission indicates the presence of newly formed dust, or pre-existing dust heated by the radiation from the supernovae or circumstellar medium interactions. We carried out a systematic analysis of the supernova host environments and their dust properties, for understanding the dust-veiled exploding stars, and whether such an intense dust production process is associated with their local environments. We find that dusty supernovae prefer the locations with higher EW(H{\alpha}), lower metallicity, and heavier host extinctions compared to typical SN types, and they show the same increasing sequence in the values of EW(H{\alpha}) and oxygen abundance from hydrogen-rich, type IIn and hydrogen-poor dusty supernovae. These differences in environmental properties of different dusty SN types indicate the diversity of their progenitors. We also found that one marginal correlation is a negative correlation between the SN dust mass and star formation rate. This means that SNe would be more mid-IR luminous and more dust-rich at the region with lower star formation rate. However, the SN dust mass show no correlation with the metallicity and the host extinction, which were thought to be key factors affecting the mass-loss history of progenitors and the CSM environment of SNe. Therefore, the dust formation process in SNe might be insensitive to metallicity and the dust condition of their host environments.

astro-ph.HE↗

arXiv2026

Cold-Start Personalization via Training-Free Priors from Structured World Models

Cold-start personalization requires inferring user preferences through interaction when no user-specific historical data is available. The core challenge is a routing problem: each task admits dozens of preference dimensions, yet individual users care about only a few, and which ones matter depends on who is asking. With a limited question budget, asking without structure will miss the dimensions that matter. Reinforcement learning is the natural formulation, but in multi-turn settings its terminal reward fails to exploit the factored, per-criterion structure of preference data, and in practice learned policies collapse to static question sequences that ignore user responses. We propose decomposing cold-start elicitation into offline structure learning and online Bayesian inference. Pep (Preference Elicitation with Priors) learns a structured world model of preference correlations offline from complete profiles, then performs training-free Bayesian inference online to select informative questions and predict complete preference profiles, including dimensions never asked about. The framework is modular across downstream solvers and requires only simple belief models. Across medical, mathematical, social, and commonsense reasoning, Pep achieves 80.8% alignment between generated responses and users' stated preferences versus 68.5% for RL, with 3-5x fewer interactions. When two users give different answers to the same question, Pep changes its follow-up 39-62% of the time versus 0-28% for RL. It does so with ~10K parameters versus 8B for RL, showing that the bottleneck in cold-start elicitation is the capability to exploit the factored structure of preference data.

cs.CL↗

arXiv2025

Smoothing DiLoCo with Primal Averaging for Faster Training of LLMs

We propose Generalized Primal Averaging (GPA), an extension of Nesterov's method that unifies and generalizes recent averaging-based optimizers like single-worker DiLoCo and Schedule-Free, within a non-distributed setting. While DiLoCo relies on a memory-intensive two-loop structure to periodically aggregate pseudo-gradients using Nesterov momentum, GPA eliminates this complexity by decoupling Nesterov's interpolation constants to enable smooth iterate averaging at every step. Structurally, GPA resembles Schedule-Free but replaces uniform averaging with exponential moving averaging. Empirically, GPA consistently outperforms single-worker DiLoCo and AdamW with reduced memory overhead. GPA achieves speedups of 8.71%, 10.13%, and 9.58% over the AdamW baseline in terms of steps to reach target validation loss for Llama-160M, 1B, and 8B models, respectively. Similarly, on the ImageNet ViT workload, GPA achieves speedups of 7% and 25.5% in the small and large batch settings respectively. Furthermore, we prove that for any base optimizer with $O(\sqrt{T})$ regret, where $T$ is the number of iterations, GPA matches or exceeds the original convergence guarantees depending on the interpolation constants.

cs.LG↗

arXiv2025

Bregman Douglas-Rachford Splitting Method

In this paper, we propose the Bregman Douglas-Rachford splitting (BDRS) method and its variant Bregman Peaceman-Rachford splitting method for solving maximal monotone inclusion problem. We show that BDRS is equivalent to a Bregman alternating direction method of multipliers (ADMM) when applied to the dual of the problem. A special case of the Bregman ADMM is an alternating direction version of the exponential multiplier method. To the best of our knowledge, algorithms proposed in this paper are new to the literature. We also discuss how to use our algorithms to solve the discrete optimal transport (OT) problem. We prove the convergence of the algorithms under certain assumptions, though we point out that one assumption does not apply to the OT problem.

math.OC↗

arXiv2025

Quantization through Piecewise-Affine Regularization: Optimization and Statistical Guarantees

Optimization problems over discrete or quantized variables are very challenging in general due to the combinatorial nature of their search space. Piecewise-affine regularization (PAR) provides a flexible modeling and computational framework for quantization based on continuous optimization. In this work, we focus on the setting of supervised learning and investigate the theoretical foundations of PAR from optimization and statistical perspectives. First, we show that in the overparameterized regime, where the number of parameters exceeds the number of samples, every critical point of the PAR-regularized loss function exhibits a high degree of quantization. Second, we derive closed-form proximal mappings for various (convex, quasi-convex, and non-convex) PARs and show how to solve PAR-regularized problems using the proximal gradient method, its accelerated variant, and the Alternating Direction Method of Multipliers. Third, we study statistical guarantees of PAR-regularized linear regression problems; specifically, we can approximate classical formulations of $\ell_1$-, squared $\ell_2$-, and nonconvex regularizations using PAR and obtain similar statistical guarantees with quantized solutions.

cs.LG↗

arXiv2025

Stochastic Approximation with Block Coordinate Optimal Stepsizes

We consider stochastic approximation with block-coordinate stepsizes and propose adaptive stepsize rules that aim to minimize the expected distance from the next iterate to an (unknown) target point. These stepsize rules employ online estimates of the second moment of the search direction along each block coordinate. The popular Adam algorithm can be interpreted as a variant with a specific estimator. By leveraging a simple conditional estimator, we derive a new method that obtains competitive performance against Adam but requires less memory and fewer hyper-parameters. We prove that this family of methods converges almost surely to a small neighborhood of the target point, and the radius of the neighborhood depends on the bias and variance of the second-moment estimator. Our analysis relies on a simple aiming condition that assumes neither convexity nor smoothness, thus has broad applicability.

math.OC↗

arXiv2025

Exploration from a Primal-Dual Lens: Value-Incentivized Actor-Critic Methods for Sample-Efficient Online RL

Online reinforcement learning (RL) with complex function approximations such as transformers and deep neural networks plays a significant role in the modern practice of artificial intelligence. Despite its popularity and importance, balancing the fundamental trade-off between exploration and exploitation remains a long-standing challenge; in particular, we are still in lack of efficient and practical schemes that are backed by theoretical performance guarantees. Motivated by recent developments in exploration via optimistic regularization, this paper provides an interpretation of the principle of optimism through the lens of primal-dual optimization. From this fresh perspective, we set forth a new value-incentivized actor-critic (VAC) method, which optimizes a single easy-to-optimize objective integrating exploration and exploitation -- it promotes state-action and policy estimates that are both consistent with collected data transitions and result in higher value functions. Theoretically, the proposed VAC method has near-optimal regret guarantees under linear Markov decision processes (MDPs) in both finite-horizon and infinite-horizon settings, which can be extended to the general function approximation setting under appropriate assumptions.

cs.LG↗

arXiv2025

Properties of the cores and filaments in the Ophiuchus molecular cloud and its L1688 hub-filament system

Analyzing filaments and cores in molecular clouds is key to understanding galactic star formation and its environmental dependence. This paper studies the properties and distribution of dense cores and filaments in the Ophiuchus molecular cloud, with a focus on the L1688 hub-filament system (HFS) and its star formation potential. We extracted sources and filaments from Herschel images and a 13.5 arcsec resolution surface density map using the getsf method, identified prestellar cores among the extracted sources, evaluated core mass segregation, and constructed the core mass function (CMF). We derived properties of the filaments from their radial surface density profiles, constructed the filament linear density function (FLDF), and assessed the mass distribution in the L1688 HFS to estimate the core and filament formation efficiencies (CFE, FFE). The results suggest that the filament-dominated core formation is a key mechanism in star formation within the system.

astro-ph.GA↗

arXiv2025

Fast Clifford Neural Layers

Clifford Neural Layers improve PDE modeling by introducing Clifford Algebra into neural networks. In this project we focus on optimizing the inference of 2/3D Clifford convolutional layers and multivector activation layers for one core CPU performance. Overall, by testing on a real network block involving Clifford convolutional layers and multivector activation layers, we observe that our implementation is 30% faster than standard PyTorch implementation in relatively large data + network size (>L2 cache). We open source our code base at https://github.com/egretwAlker/c-opt-clifford-layers

cs.LG↗

arXiv2025

Very Late-Time JWST and Keck Spectra of the Oxygen-Rich Supernova 1995N

We present new {\it JWST}/MIRI MRS and Keck spectra of SN 1995N obtained in 2022--2023, more than 10,000 days after the supernova (SN) explosion. These spectra are among the latest direct detections of a core-collapse SN, both through emission lines in the optical and thermal continuum from infrared dust emission. The new infrared data show that dust heating from radiation produced by the ejecta interacting with circumstellar matter is still present, but greatly reduced from when SN 1995N was observed by the {\it Spitzer Space Telescope} and {\it WISE} in 2009/2010 and 2018, when the dust mass was estimated to be 0.4 M(Sun). New radiative-transfer modeling suggests that the dust mass and grain size may have increased between 2010 and 2023. The new data can alternatively be well fit with a dust mass of 0.4 M(Sun) and a much reduced heating source luminosity. The new late-time spectra show unusually strong oxygen forbidden lines, stronger than the H-alpha emission. This indicates that SN 1995N may have exploded as a stripped-envelope SN which then interacted with a massive H-rich circumstellar shell, changing it from intrinsically Type Ib/c to Type IIn. The late-time spectrum results when the reverse shock begins to excite the inner H-poor, O-rich ejecta. This change in the spectrum is rarely seen, but marks the start of the transition from SN to SN remnant.

astro-ph.SR↗

arXiv2025

LoRe: Personalizing LLMs via Low-Rank Reward Modeling

Personalizing large language models (LLMs) to accommodate diverse user preferences is essential for enhancing alignment and user satisfaction. Traditional reinforcement learning from human feedback (RLHF) approaches often rely on monolithic value representations, limiting their ability to adapt to individual preferences. We introduce a novel framework that leverages low-rank preference modeling to efficiently learn and generalize user-specific reward functions. By representing reward functions in a low-dimensional subspace and modeling individual preferences as weighted combinations of shared basis functions, our approach avoids rigid user categorization while enabling scalability and few-shot adaptation. We validate our method on multiple preference datasets, demonstrating superior generalization to unseen users and improved accuracy in preference prediction tasks.

cs.LG↗

arXiv2025

Salient Object Detection in Traffic Scene through the TSOD10K Dataset

Traffic Salient Object Detection (TSOD) aims to segment the objects critical to driving safety by combining semantic (e.g., collision risks) and visual saliency. Unlike SOD in natural scene images (NSI-SOD), which prioritizes visually distinctive regions, TSOD emphasizes the objects that demand immediate driver attention due to their semantic impact, even with low visual contrast. This dual criterion, i.e., bridging perception and contextual risk, re-defines saliency for autonomous and assisted driving systems. To address the lack of task-specific benchmarks, we collect the first large-scale TSOD dataset with pixel-wise saliency annotations, named TSOD10K. TSOD10K covers the diverse object categories in various real-world traffic scenes under various challenging weather/illumination variations (e.g., fog, snowstorms, low-contrast, and low-light). Methodologically, we propose a Mamba-based TSOD model, termed Tramba. Considering the challenge of distinguishing inconspicuous visual information from complex traffic backgrounds, Tramba introduces a novel Dual-Frequency Visual State Space module equipped with shifted window partitioning and dilated scanning to enhance the perception of fine details and global structure by hierarchically decomposing high/low-frequency components. To emphasize critical regions in traffic scenes, we propose a traffic-oriented Helix 2D-Selective-Scan (Helix-SS2D) mechanism that injects driving attention priors while effectively capturing global multi-direction spatial dependencies. We establish a comprehensive benchmark by evaluating Tramba and 22 existing NSI-SOD models on TSOD10K, demonstrating Tramba's superiority. Our research establishes the first foundation for safety-aware saliency analysis in intelligent transportation systems.

cs.CV↗

arXiv2025

PARQ: Piecewise-Affine Regularized Quantization

We develop a principled method for quantization-aware training (QAT) of large-scale machine learning models. Specifically, we show that convex, piecewise-affine regularization (PAR) can effectively induce the model parameters to cluster towards discrete values. We minimize PAR-regularized loss functions using an aggregate proximal stochastic gradient method (AProx) and prove that it has last-iterate convergence. Our approach provides an interpretation of the straight-through estimator (STE), a widely used heuristic for QAT, as the asymptotic form of PARQ. We conduct experiments to demonstrate that PARQ obtains competitive performance on convolution- and transformer-based vision tasks.

cs.LG↗

arXiv2025

JWST/MIRI detects the dusty SN1993J about 30 years after explosion

Core-collapse supernovae (CCSNe) have long been considered to contribute significantly to the cosmic dust budget. New dust cools quickly and is therefore detectable at mid-infrared (mid-IR) wavelengths. However, before the era of the James Webb Space Telescope (JWST), direct observational evidence for dust condensation was found in only a handful of nearby CCSNe, and dust masses (~10$^{-2}-10^{-3} M_{\odot}$, generally limited to <5 yr and to >500K temperatures) have been 2-3 orders of magnitude smaller than either theoretical predictions or dust amounts found by far-IR/submm observations of Galactic SN remnants and in the very nearby SN 1987A. The combined angular resolution and mid-IR sensitivity of JWST finally allow us to reveal hidden cool (~100-200K) dust reservoirs in extragalactic SNe beyond SN 1987A. Our team received JWST/MIRI time for studying a larger sample of CCSNe to fill the currently existing gap in their dust formation histories. The first observed target of this program is the well-known Type IIb SN~1993J appeared in M81. We generated its spectral energy distribution (SED) from the current JWST/MIRI F770W, F1000W, F1500W, and F2100W fluxes. We fit single- and two-component silicate and carbonaceous dust models to the SED. We found that SN 1993J still contains a significant amount (~0.01 $M_{\odot}$) of dust ~30 yr after explosion. Comparing these results to those of the analysis of earlier {Spitzer Space Telescope data, we see a similar amount of dust now that was detected ~15-20 yr ago, but at a lower temperature. We also find residual background emission near the SN site (after point-spread-function subtraction on the JWST/MIRI images) that may plausibly be attributed to an IR echo from more distant interstellar dust grains heated by the SN shock-breakout luminosity or ongoing star formation in the local environment.

astro-ph.SR↗

arXiv2025

Incentivize without Bonus: Provably Efficient Model-based Online Multi-agent RL for Markov Games

Multi-agent reinforcement learning (MARL) lies at the heart of a plethora of applications involving the interaction of a group of agents in a shared unknown environment. A prominent framework for studying MARL is Markov games, with the goal of finding various notions of equilibria in a sample-efficient manner, such as the Nash equilibrium (NE) and the coarse correlated equilibrium (CCE). However, existing sample-efficient approaches either require tailored uncertainty estimation under function approximation, or careful coordination of the players. In this paper, we propose a novel model-based algorithm, called VMG, that incentivizes exploration via biasing the empirical estimate of the model parameters towards those with a higher collective best-response values of all the players when fixing the other players' policies, thus encouraging the policy to deviate from its current equilibrium for more exploration. VMG is oblivious to different forms of function approximation, and permits simultaneous and uncoupled policy updates of all players. Theoretically, we also establish that VMG achieves a near-optimal regret for finding both the NEs of two-player zero-sum Markov games and CCEs of multi-player general-sum Markov games under linear function approximation in an online environment, which nearly match their counterparts with sophisticated uncertainty quantification.

cs.LG↗

arXiv2025

ParetoQ: Improving Scaling Laws in Extremely Low-bit LLM Quantization

The optimal bit-width for achieving the best trade-off between quantized model size and accuracy has been a subject of ongoing debate. While some advocate for 4-bit quantization, others propose that 1.58-bit offers superior results. However, the lack of a cohesive framework for different bits has left such conclusions relatively tenuous. We present ParetoQ, the first unified framework that facilitates rigorous comparisons across 1-bit, 1.58-bit, 2-bit, 3-bit, and 4-bit quantization settings. Our findings reveal a notable learning transition between 2 and 3 bits: For 3-bits and above, the fine-tuned models stay close to their original pre-trained distributions, whereas for learning 2-bit networks or below, the representations change drastically. By optimizing training schemes and refining quantization functions, ParetoQ surpasses all previous methods tailored to specific bit widths. Remarkably, our ParetoQ ternary 600M-parameter model even outperforms the previous SoTA ternary 3B-parameter model in accuracy, using only one-fifth of the parameters. Extensive experimentation shows that ternary, 2-bit, and 3-bit quantization maintains comparable performance in the size-accuracy trade-off and generally exceeds 4-bit and binary quantization. Considering hardware constraints, 2-bit quantization offers promising potential for memory reduction and speedup.

cs.LG↗

arXiv2025

SHYI: Action Support for Contrastive Learning in High-Fidelity Text-to-Image Generation

In this project, we address the issue of infidelity in text-to-image generation, particularly for actions involving multiple objects. For this we build on top of the CONFORM framework which uses Contrastive Learning to improve the accuracy of the generated image for multiple objects. However the depiction of actions which involves multiple different object has still large room for improvement. To improve, we employ semantically hypergraphic contrastive adjacency learning, a comprehension of enhanced contrastive structure and "contrast but link" technique. We further amend Stable Diffusion's understanding of actions by InteractDiffusion. As evaluation metrics we use image-text similarity CLIP and TIFA. In addition, we conducted a user study. Our method shows promising results even with verbs that Stable Diffusion understands mediocrely. We then provide future directions by analyzing the results. Our codebase can be found on polybox under the link: https://polybox.ethz.ch/index.php/s/dJm3SWyRohUrFxn

cs.CV↗

arXiv2025

DiffCL: A Diffusion-Based Contrastive Learning Framework with Semantic Alignment for Multimodal Recommendations

Multimodal recommendation systems integrate diverse multimodal information into the feature representations of both items and users, thereby enabling a more comprehensive modeling of user preferences. However, existing methods are hindered by data sparsity and the inherent noise within multimodal data, which impedes the accurate capture of users' interest preferences. Additionally, discrepancies in the semantic representations of items across different modalities can adversely impact the prediction accuracy of recommendation models. To address these challenges, we introduce a novel diffusion-based contrastive learning framework (DiffCL) for multimodal recommendation. DiffCL employs a diffusion model to generate contrastive views that effectively mitigate the impact of noise during the contrastive learning phase. Furthermore, it improves semantic consistency across modalities by aligning distinct visual and textual semantic information through stable ID embeddings. Finally, the introduction of the Item-Item Graph enhances multimodal feature representations, thereby alleviating the adverse effects of data sparsity on the overall system performance. We conduct extensive experiments on three public datasets, and the results demonstrate the superiority and effectiveness of the DiffCL.

cs.MM↗