arXiv subjects

Jiale Chen

Publications and source records attributed to Jiale Chen.

At least 37 records · Page 2Linked to original sources

arXiv2026

Scaling Reasoning Tokens via RL and Parallel Thinking: Evidence From Competitive Programming

We study how to scale reasoning token budgets for competitive programming through two complementary approaches: training-time reinforcement learning (RL) and test-time parallel thinking. During RL training, we observe an approximately log-linear relationship between validation accuracy and the average number of generated reasoning tokens over successive checkpoints, and show two ways to shift this training trajectory: verification RL warmup raises the starting point, while randomized clipping produces a steeper trend in the observed regime. As scaling single-generation reasoning during RL quickly becomes expensive under full attention, we introduce a multi-round parallel thinking pipeline that distributes the token budget across threads and rounds of generation, verification, and refinement. We train the model end-to-end on this pipeline to match the training objective to the test-time structure. Starting from Seed-OSS-36B, the full system with 16 threads and 16 rounds per thread matches the underlying RL model's oracle pass@16 at pass@1 using 7.6 million tokens per problem on average, and surpasses GPT-5-high on 456 hard competitive programming problems from AetherCode.

cs.CL↗

arXiv2026

What, Whether and How? Unveiling Process Reward Models for Thinking with Images Reasoning

The rapid advancement of Large Vision Language Models (LVLMs) has demonstrated excellent abilities in various visual tasks. Building upon these developments, the thinking with images paradigm has emerged, enabling models to dynamically edit and re-encode visual information at each reasoning step, mirroring human visual processing. However, this paradigm introduces significant challenges as diverse errors may occur during reasoning processes. This necessitates Process Reward Models (PRMs) for distinguishing positive and negative reasoning steps, yet existing benchmarks for PRMs are predominantly text-centric and lack comprehensive assessment under this paradigm. To address these gaps, this work introduces the first comprehensive benchmark specifically designed for evaluating PRMs under the thinking with images paradigm. Our main contributions are: (1) Through extensive analysis of reasoning trajectories and guided search experiments with PRMs, we define 7 fine-grained error types and demonstrate both the necessity for specialized PRMs and the potential for improvement. (2) We construct a comprehensive benchmark comprising 1,206 manually annotated thinking with images reasoning trajectories spanning 4 categories and 16 subcategories for fine-grained evaluation of PRMs. (3) Our experimental analysis reveals that current LVLMs fall short as effective PRMs, exhibiting limited capabilities in visual reasoning process evaluation with significant performance disparities across error types, positive evaluation bias, and sensitivity to reasoning step positions. These findings demonstrate the effectiveness of our benchmark and establish crucial foundations for advancing PRMs in LVLMs.

cs.CV↗

arXiv2026

Embedding theorems and integration operators on Hardy--Carleson type tent spaces induced by doubling weights

This paper develops the function and operator theory of Hardy--Carleson--type analytic tent spaces $AT_q^\infty(ω)$ induced by radial weights $ω$ satisfying a two-sided doubling condition. We first characterize the positive Borel measures $μ$ for which the embedding from $AT_p^\infty(ω)$ into the tent space $T_q^\infty(μ)$ is bounded for all $0 < p, q < \infty$. A Littlewood--Paley formula for $AT_q^\infty(ω)$ is then established. Using these results, we give a complete characterization of the boundedness (compactness) of Volterra-type integration operators between $AT_p^\infty(ω)$ and $AT_q^\infty(ω)$.

math.CV↗

arXiv2026

Quartet: Native FP4 Training Can Be Optimal for Large Language Models

Training large language models (LLMs) models directly in low-precision offers a way to address computational costs by improving both throughput and energy efficiency. For those purposes, NVIDIA's recent Blackwell architecture facilitates very low-precision operations using FP4 variants. Yet, current algorithms for training LLMs in FP4 precision face significant accuracy degradation and often rely on mixed-precision fallbacks. In this paper, we investigate hardware-supported FP4 training and introduce a new approach for accurate, end-to-end FP4 training with all the major computations (i.e., linear layers) in low precision. Through extensive evaluations on Llama-type models, we reveal a new low-precision scaling law that quantifies performance trade-offs across bit-widths and training setups. Guided by this investigation, we design an "optimal" technique in terms of accuracy-vs-computation, called Quartet. We implement Quartet using optimized CUDA kernels tailored for Blackwell, demonstrating that fully FP4-based training is a competitive alternative to FP16 half-precision and to FP8 training. Our code is available at https://github.com/IST-DASLab/Quartet.

cs.LG↗

arXiv2025

A Comprehensive Study of Bugs in Modern Distributed Deep Learning Systems

In today's data-driven era, deep learning is vital for processing massive datasets, yet single-device training is constrained by computational and memory limits. Distributed deep learning overcomes these challenges by leveraging multiple GPUs or machines in parallel. While general-purpose frameworks (e.g., TensorFlow and PyTorch) provide distributed capabilities, these are often add-on features that demand significant manual effort for advanced parallelism, underscoring the need for specialized frameworks. This study conducts the first large-scale empirical analysis of practitioner challenges in dedicated distributed frameworks. We examine 849 real-world issues from DeepSpeed, Megatron-LM, and Colossal-AI and construct a taxonomy of 34 bug symptoms, 28 root causes, and 6 fix patterns. Crucially, we establish explicit mappings between symptoms, causes, and fixes across distributed training stages, enabling a systematic understanding of how issues emerge and are resolved. Our results show that 45.1\% of bug symptoms are unique to distributed frameworks, with setup failures, memory issues, and performance anomalies being the most prevalent. Moreover, 95\% of issues in the communication setup stage occur exclusively in distributed contexts. We also find over 60\% of cases can be resolved through version and dependency management, and distributed feature, API, and communication tuning. Based on these findings, we provide actionable implications.

cs.SE↗

arXiv2025

Toeplitz operators on weighted Fock spaces with $A_{\infty}$-type weights

By establishing some reproducing kernel estimates, we characterize the bounded, compact and Schatten $p$-class Toeplitz operators with positive measure symbols on the weighted Fock space $F^2_{α,w}$ for $p\geq1$, where $w$ is a weight on the complex plane satisfying an $A_{\infty}$-type condition. Applications to Volterra operators and weighted composition operators are given.

math.FA↗

arXiv2025

Weak-type bounds for the Bergman projection with Bekollé-Bonami weights

We establish weighted weak-type bounds for the Bergman projection with respect to Bekollé-Bonami characteristics. We present two proofs of an improved quantitative weak-type $(1,1)$ estimate, as well as sharp weak-type $(p,p)$ bounds for $p>1$ and mixed weighted weak-type $(1,1)$ inequalities. Our results, which hold for a wide class of simple domains in $\mathbb{C}^n$, are new even in the classical settings of the upper half-plane and the unit disk.

math.FA↗

arXiv2025

HSKBenchmark: Modeling and Benchmarking Chinese Second Language Acquisition in Large Language Models through Curriculum Tuning

Language acquisition is vital to revealing the nature of human language intelligence and has recently emerged as a promising perspective for improving the interpretability of large language models (LLMs). However, it is ethically and practically infeasible to conduct experiments that require controlling human learners' language inputs. This poses challenges for the verifiability and scalability of language acquisition modeling, particularly in Chinese second language acquisition (SLA). While LLMs provide a controllable and reproducible alternative, a systematic benchmark to support phase-wise modeling and assessment is still lacking. In this paper, we present HSKBenchmark, the first benchmark for staged modeling and writing assessment of LLMs in Chinese SLA. It covers HSK levels 3 to 6 and includes authentic textbooks with 6.76 million tokens, 16K synthetic instruction samples, 30 test topics, and a linguistically grounded evaluation system. To simulate human learning trajectories, we introduce a curriculum-tuning framework that trains models from beginner to advanced levels. An evaluation system is created to examine level-based grammar coverage, writing errors, lexical and syntactic complexity, and holistic scoring. We also build HSKAgent, fine-tuned on 10K learner compositions. Extensive experimental results demonstrate that HSKBenchmark not only models Chinese SLA effectively, but also serves as a reliable benchmark for dynamic writing assessment in LLMs. Our fine-tuned LLMs have writing performance on par with advanced human learners and exhibit human-like acquisition characteristics. The HSKBenchmark, HSKAgent, and checkpoints serve as foundational tools and resources, with the potential to pave the way for future research on language acquisition modeling and LLMs interpretability. Code and data are publicly available at: https://github.com/CharlesYang030/HSKB.

cs.CL↗

arXiv2025

Scalable Multitemperature Free Energy Sampling of Classical Ising Spin States

Generative models have advanced significantly in sampling material systems with continuous variables, such as atomistic structures. However, their application to discrete variables, like atom types or spin states, remains underexplored. In this work, we introduce a discrete flow matching model, tailored for systems with discrete phase-space coordinates (e.g., the Ising model or a multicomponent system on a lattice). This approach enables a single model to sample free energy surfaces over a wide temperature range with minimal training overhead, and the model generation is scalable to larger lattice sizes than those in the training set. We demonstrate our approach on the 2D Ising model, showing efficient and reliable free energy sampling. These results highlight the potential of flow matching for low-cost, scalable free energy sampling in discrete systems and suggest promising extensions to alchemical degrees of freedom in crystalline materials. The codebase developed for this work is openly available at https://github.com/tuoping/alchemicalFES.

cond-mat.stat-mech↗

arXiv2025

Flow matching for reaction pathway generation

Elucidating reaction mechanisms hinges on efficiently generating transition states (TSs), products, and complete reaction networks. Recent generative models, such as diffusion models for TS sampling and sequence-based architectures for product generation, offer faster alternatives to quantum-chemistry searches. But diffusion models remain constrained by their stochastic differential equation (SDE) dynamics, which suffer from inefficiency and limited controllability. We show that flow matching, a deterministic ordinary differential (ODE) formulation, can replace SDE-based diffusion for molecular and reaction generation. We introduce MolGEN, a conditional flow-matching framework that learns an optimal transport path to transport Gaussian priors to target chemical distributions. On benchmarks used by TSDiff and OA-ReactDiff, MolGEN surpasses TS geometry accuracy and barrier-height prediction while reducing sampling to sub-second inference. MolGEN also supports open-ended product generation with competitive top-k accuracy and avoids mass/electron-balance violations common to sequence models. In a realistic test on the $γ$-ketohydroperoxide decomposition network, MolGEN yields higher fractions of valid and intended TSs with markedly fewer quantum-chemistry evaluations than string-based baselines. These results demonstrate that deterministic flow matching provides a unified, accurate, and computationally efficient foundation for molecular generative modeling, signaling that flow matching is the future for molecular generation across chemistry.

physics.chem-ph↗

arXiv2025

Fock projections on vector-valued $L^p$-spaces with matrix weights

In this paper, we characterize the $d\times d$ matrix weights $W$ on $\mathbb{C}^n$ such that the Fock projection $P_α$ is bounded on the vector-valued spaces $L^p_{α,W}(\mathbb{C}^n;\mathbb{C}^d)$ induced by $W$ and the Gaussian measures. It is proved that for $1\leq p\leq\infty$, the Fock projection $P_α$ is bounded on $L^p_{α,W}(\mathbb{C}^n;\mathbb{C}^d)$ if and only if $W$ satisfies a restricted $\mathcal{A}_p$-condition. Our result is new even in the scalar setting at the endpoint $p=\infty$.

math.FA↗

arXiv2025

From Unweighted to Weighted Dynamic Matching in Non-Bipartite Graphs: A Low-Loss Reduction

We study the approximate maximum weight matching (MWM) problem in a fully dynamic graph subject to edge insertions and deletions. We design meta-algorithms that reduce the problem to the unweighted approximate maximum cardinality matching (MCM) problem. Despite recent progress on bipartite graphs -- Bernstein-Dudeja-Langley (STOC 2021) and Bernstein-Chen-Dudeja-Langley-Sidford-Tu (SODA 2025) -- the only previous meta-algorithm that applied to non-bipartite graphs suffered a $\frac{1}{2}$ approximation loss (Stubbs-Williams, ITCS 2017). We significantly close the weighted-and-unweighted gap by showing the first low-loss reduction that transforms any fully dynamic $(1-\varepsilon)$-approximate MCM algorithm on bipartite graphs into a fully dynamic $(1-\varepsilon)$-approximate MWM algorithm on general (not necessarily bipartite) graphs, with only a $\mathrm{poly}(\log n/\varepsilon)$ overhead in the update time. Central to our approach is a new primal-dual framework that reduces the computation of an approximate MWM in general graphs to a sequence of approximate induced matching queries on an auxiliary bipartite extension. In addition, we give the first conditional lower bound on approximate partially dynamic matching with worst-case update time.

cs.DS↗

arXiv2025

Empathy Omni: Enabling Empathetic Speech Response Generation through Large Language Models

With the development of speech large language models (speech LLMs), users can now interact directly with assistants via speech. However, most existing models only convert response content into speech without fully capturing the rich emotional cues in user queries, where the same sentence may convey different meanings depending on the expression. Emotional understanding is thus essential for improving human-machine interaction. Most empathetic speech LLMs rely on massive datasets, demanding high computational cost. A key challenge is to build models that generate empathetic responses with limited data and without large-scale training. To this end, we propose Emotion Omni, a model that understands emotional content in user speech and generates empathetic responses. We further developed a data pipeline to construct a 200k emotional dialogue dataset supporting empathetic speech assistants. Experiments show that Emotion Omni achieves comparable instruction-following ability without large-scale pretraining, while surpassing existing models in speech quality (UTMOS:4.41) and empathy (Emotion GPT Score: 3.97). These results confirm its improvements in both speech fidelity and emotional expressiveness. Demos are available at https://w311411.github.io/omni_demo/.

cs.CL↗

arXiv2025

T-Stars-Poster: A Framework for Product-Centric Advertising Image Design

Creating advertising images is often a labor-intensive and time-consuming process. Can we automatically generate such images using basic product information like a product foreground image, taglines, and a target size? Existing methods mainly focus on parts of the problem and lack a comprehensive solution. To bridge this gap, we propose a novel product-centric framework for advertising image design called T-Stars-Poster. It consists of four sequential stages to highlight product foregrounds and taglines while achieving overall image aesthetics: prompt generation, layout generation, background image generation, and graphics rendering. Different expert models are designed and trained for the first three stages: First, a visual language model (VLM) generates background prompts that match the products. Next, a VLM-based layout generation model arranges the placement of product foregrounds, graphic elements (taglines and decorative underlays), and various nongraphic elements (objects from the background prompt). Following this, an SDXL-based model can simultaneously accept prompts, layouts, and foreground controls to generate images. To support T-Stars-Poster, we create two corresponding datasets with over 50,000 labeled images. Extensive experiments and online A/B tests demonstrate that T-Stars-Poster can produce more visually appealing advertising images.

cs.CV↗

arXiv2025

CTR-Driven Ad Text Generation via Online Feedback Preference Optimization

Advertising text plays a critical role in determining click-through rates (CTR) in online advertising. Large Language Models (LLMs) offer significant efficiency advantages over manual ad text creation. However, LLM-generated ad texts do not guarantee higher CTR performance compared to human-crafted texts, revealing a gap between generation quality and online performance of ad texts. In this work, we propose a novel ad text generation method which optimizes for CTR through preference optimization from online feedback. Our approach adopts an innovative two-stage framework: (1) diverse ad text sampling via one-shot in-context learning, using retrieval-augmented generation (RAG) to provide exemplars with chain-of-thought (CoT) reasoning; (2) CTR-driven preference optimization from online feedback, which weighs preference pairs according to their CTR gains and confidence levels. Through our method, the resulting model enables end-to-end generation of high-CTR ad texts. Extensive experiments have demonstrated the effectiveness of our method in both offline and online metrics. Notably, we have applied our method on a large-scale online shopping platform and achieved significant CTR improvements, showcasing its strong applicability and effectiveness in advertising systems.

cs.IR↗

arXiv2025

Self-Supervised Contrastive Learning for Multi-Label Images

Self-supervised learning (SSL) has demonstrated its effectiveness in learning representations through comparison methods that align with human intuition. However, mainstream SSL methods heavily rely on high body datasets with single label, such as ImageNet, resulting in intolerable pre-training overhead. Besides, more general multi-label images are frequently overlooked in SSL, despite their potential for richer semantic information and broader applicability in downstream scenarios. Therefore, we tailor the mainstream SSL approach to guarantee excellent representation learning capabilities using fewer multi-label images. Firstly, we propose a block-wise augmentation module aimed at extracting additional potential positive view pairs from multi-label images. Subsequently, an image-aware contrastive loss is devised to establish connections between these views, thereby facilitating the extraction of semantically consistent representations. Comprehensive linear fine-tuning and transfer learning validate the competitiveness of our approach despite challenging sample quality and quantity.

cs.CV↗

arXiv2025

QuEST: Stable Training of LLMs with 1-Bit Weights and Activations

One approach to reducing the massive costs of large language models (LLMs) is the use of quantized or sparse representations for training or deployment. While post-training compression methods are very popular, the question of obtaining even more accurate compressed models by directly training over such representations, i.e., Quantization-Aware Training (QAT), is still open: for example, a recent study (arXiv:2411.04330) put the "optimal" bit-width at which models can be trained using QAT, while staying accuracy-competitive with standard FP16/BF16 precision, at 8-bits weights and activations. We advance this state-of-the-art via a new method called QuEST, for which we demonstrate optimality at 4-bits and stable convergence as low as 1-bit weights and activations. QuEST achieves this by improving two key aspects of QAT methods: (1) accurate and fast quantization of the (continuous) distributions of weights and activations via Hadamard normalization and MSE-optimal fitting; (2) a new trust gradient estimator based on the idea of explicitly minimizing the error between the noisy gradient computed over quantized states and the "true" (but unknown) full-precision gradient. Experiments on Llama-type architectures show that QuEST induces stable scaling laws across the entire range of hardware-supported precisions, and can be extended to sparse representations. We provide GPU kernel support showing that models produced by QuEST can be executed efficiently. Our code is available at https://github.com/IST-DASLab/QuEST.

cs.LG↗

arXiv2025

Ferromagnetic Two-dimensional Electron Gases with Magnetic Doping and Proximity Effects

The advent of magnetic two-dimensional electron gases (2DEGs) at oxide interfaces has provided new opportunities in the field of spintronics. The enhancement of magnetism in 2DEGs at oxide interfaces continues to be a significant challenge, as exemplified by the relatively weak magnetism observed in the classical LaAlO3/SrTiO3 interface. Here, we present ferromagnetic (FM) 2DEGs at the interface fabricated between the FM insulator EuTiO3 (ETO) and the strong spin-orbit coupled (SOC) perovskite insulator KTaO3 (KTO). With the combined effects of magnetic atom doping and magnetic proximity from ETO films, the coercive field of 2DEGs can be significantly enhanced. Magnetoresistance (MR) curve with a high coercive field of 1000 Oe has been observed, in conjunction with a temperature-dependent unambiguous hysteresis loop in the anomalous Hall effect (AHE). Furthermore, within the 2DEGs, we have identified a synergetic interplay between magnetic scattering and the weak antilocalization (WAL) effect on transport. This study provides fresh insights into the formation of FM 2DEGs at ETO/KTO interfaces, and introduce an innovative pathway for creating high-performance magnetic 2DEGs at oxide interfaces.

cond-mat.mtrl-sci↗