Xumeng Wen

Xumeng Wen, Han Zhang, Shun Zheng et al.

Tabular data is foundational to predictive modeling in various crucial industries, including healthcare, finance, retail, sustainability, etc. Despite the progress made in specialized models, there is an increasing demand for universal models that can transfer knowledge, generalize from limited data, and follow human instructions. These are challenges that current tabular deep learning approaches have not fully tackled. Here we introduce Generative Tabular Learning (GTL), a novel framework that integrates the advanced functionalities of large language models (LLMs)-such as prompt-based zero-shot generalization and in-context learning-into tabular deep learning. GTL capitalizes on the pre-training of LLMs on diverse tabular data, enhancing their understanding of domain-specific knowledge, numerical sequences, and statistical dependencies critical for accurate predictions. Our empirical study spans 384 public datasets, rigorously analyzing GTL's convergence and scaling behaviors and assessing the impact of varied data templates. The GTL-enhanced LLaMA-2 model demonstrates superior zero-shot and in-context learning capabilities across numerous classification and regression tasks. Notably, it achieves this without fine-tuning, outperforming traditional methods and rivaling state-of-the-art models like GPT-4 in certain cases. Through GTL, we not only foster a deeper integration of LLMs' sophisticated abilities into tabular data comprehension and application but also offer a new training resource and a test bed for LLMs to enhance their ability to comprehend tabular data. To facilitate reproducible research, we release our code, data, and model checkpoints at https://github.com/microsoft/Industrial-Foundation-Models.

Chenguo Lin, Xumeng Wen, Wei Cao et al.

Recent research on time-series self-supervised models shows great promise in learning semantic representations. However, it has been limited to small-scale datasets, e.g., thousands of temporal sequences. In this work, we make key technical contributions that are tailored to the numerical properties of time-series data and allow the model to scale to large datasets, e.g., millions of temporal sequences. We adopt the Transformer architecture by first partitioning the input into non-overlapping windows. Each window is then characterized by its normalized shape and two scalar values denoting the mean and standard deviation within each window. To embed scalar values that may possess arbitrary numerical amplitudes in a high-dimensional space, we propose a numerically multi-scaled embedding module enumerating all possible numerical scales for the scalars. The model undergoes pretraining with a simple contrastive objective on a large-scale dataset over a million sequences collected by merging existing public data. We study its transfer performance on a number of univariate and multivariate classification tasks, few shot learning, unsupervised clustering and anomaly detection benchmarks. Our method exhibits remarkable improvement against previous pretraining approaches and establishes the new state of the art, even compared with domain-specific non-learning-based methods. Code is available at: \url{https://github.com/chenguolin/NuTime}.

Jiawen Zhang, Xumeng Wen, Zhenwei Zhang et al.

Delivering precise point and distributional forecasts across a spectrum of prediction horizons represents a significant and enduring challenge in the application of time-series forecasting within various industries. Prior research on developing deep learning models for time-series forecasting has often concentrated on isolated aspects, such as long-term point forecasting or short-term probabilistic estimations. This narrow focus may result in skewed methodological choices and hinder the adaptability of these models to uncharted scenarios. While there is a rising trend in developing universal forecasting models, a thorough understanding of their advantages and drawbacks, especially regarding essential forecasting needs like point and distributional forecasts across short and long horizons, is still lacking. In this paper, we present ProbTS, a benchmark tool designed as a unified platform to evaluate these fundamental forecasting needs and to conduct a rigorous comparative analysis of numerous cutting-edge studies from recent years. We dissect the distinctive data characteristics arising from disparate forecasting requirements and elucidate how these characteristics can skew methodological preferences in typical research trajectories, which often fail to fully accommodate essential forecasting needs. Building on this, we examine the latest models for universal time-series forecasting and discover that our analyses of methodological strengths and weaknesses are also applicable to these universal models. Finally, we outline the limitations inherent in current research and underscore several avenues for future exploration.

Xumeng Wen, Zihan Liu, Shun Zheng et al.

Recent advancements in long chain-of-thought (CoT) reasoning, particularly through the Group Relative Policy Optimization algorithm used by DeepSeek-R1, have led to significant interest in the potential of Reinforcement Learning with Verifiable Rewards (RLVR) for Large Language Models (LLMs). While RLVR promises to improve reasoning by allowing models to learn from free exploration, there remains debate over whether it truly enhances reasoning abilities or simply boosts sampling efficiency. This paper systematically investigates the impact of RLVR on LLM reasoning. We revisit Pass@K experiments and demonstrate that RLVR can extend the reasoning boundary for both mathematical and coding tasks. This is supported by our introduction of a novel evaluation metric, CoT-Pass@K, which captures reasoning success by accounting for both the final answer and intermediate reasoning steps. Furthermore, we present a theoretical framework explaining RLVR's incentive mechanism, demonstrating how it can encourage correct reasoning even when rewards are based solely on answer correctness. Our analysis of RLVR's training dynamics reveals that it incentivizes correct reasoning early in the process, with substantial improvements in reasoning quality confirmed through extensive evaluations. These findings provide strong evidence of RLVR's potential to enhance LLM reasoning, offering valuable insights into its mechanisms and performance improvements.

Jiawen Zhang, Shun Zheng, Xumeng Wen et al.

Numerous industrial sectors necessitate models capable of providing robust forecasts across various horizons. Despite the recent strides in crafting specific architectures for time-series forecasting and developing pre-trained universal models, a comprehensive examination of their capability in accommodating varied-horizon forecasting during inference is still lacking. This paper bridges this gap through the design and evaluation of the Elastic Time-Series Transformer (ElasTST). The ElasTST model incorporates a non-autoregressive design with placeholders and structured self-attention masks, warranting future outputs that are invariant to adjustments in inference horizons. A tunable version of rotary position embedding is also integrated into ElasTST to capture time-series-specific periods and enhance adaptability to different horizons. Additionally, ElasTST employs a multi-scale patch design, effectively integrating both fine-grained and coarse-grained information. During the training phase, ElasTST uses a horizon reweighting strategy that approximates the effect of random sampling across multiple horizons with a single fixed horizon setting. Through comprehensive experiments and comparisons with state-of-the-art time-series architectures and contemporary foundation models, we demonstrate the efficacy of ElasTST's unique design elements. Our findings position ElasTST as a robust solution for the practical necessity of varied-horizon forecasting.

Xumeng Wen, Shun Zheng, Zhen Xu et al.

Recent studies have shown that large language models (LLMs), when customized with post-training on tabular data, can acquire general tabular in-context learning (TabICL) capabilities. These models are able to transfer effectively across diverse data schemas and different task domains. However, existing LLM-based TabICL approaches are constrained to few-shot scenarios due to the sequence length limitations of LLMs, as tabular instances represented in plain text consume substantial tokens. To address this limitation and enable scalable TabICL for any data size, we propose retrieval-augmented LLMs tailored to tabular data. Our approach incorporates a customized retrieval module, combined with retrieval-guided instruction-tuning for LLMs. This enables LLMs to effectively leverage larger datasets, achieving significantly improved performance across 69 widely recognized datasets and demonstrating promising scaling behavior. Extensive comparisons with state-of-the-art tabular models reveal that, while LLM-based TabICL still lags behind well-tuned numeric models in overall performance, it uncovers powerful algorithms under limited contexts, enhances ensemble diversity, and excels on specific datasets. These unique properties underscore the potential of language as a universal and accessible interface for scalable tabular data learning.

Zihan Liu, Shun Zheng, Xumeng Wen et al.

Long-form chain-of-thought reasoning has become a cornerstone of advanced reasoning in large language models. While recent verification-refinement frameworks have enabled proprietary models to solve Olympiad-level problems, their effectiveness hinges on strong, reliable verification and correction capabilities, which remain fragile in open-weight, smaller-scale models. This work demonstrates that even with weak verification and refinement capabilities on hard tasks, the reasoning limits of such models can be substantially extended through a probabilistic paradigm we call Deep Self-Evolving Reasoning (DSER). We conceptualize iterative reasoning as a Markov chain, where each step represents a stochastic transition in the solution space. The key insight is that convergence to a correct solution is guaranteed as long as the probability of improvement marginally exceeds that of degradation. By running multiple long-horizon, self-evolving processes in parallel, DSER amplifies these small positive tendencies, enabling the model to asymptotically approach correct answers. Empirically, we apply DSER to the DeepSeek-R1-0528-Qwen3-8B model. On the challenging AIME 2024-2025 benchmark, DSER solves 5 out of 9 previously unsolvable problems and boosts overall performance, enabling this compact model to surpass the single-turn accuracy of its 600B-parameter teacher through majority voting. Beyond its immediate utility for test-time scaling, the DSER framework serves to diagnose the fundamental limitations of current open-weight reasoners. By clearly delineating their shortcomings in self-verification, refinement, and stability, our findings establish a clear research agenda for developing next-generation models with powerful, intrinsic self-evolving capabilities.

Jiawen Zhang, Zhenwei Zhang, Shun Zheng et al.

Time Series Foundation Models (TSFMs), which are pretrained on large-scale, cross-domain data and capable of zero-shot forecasting in new scenarios without further training, are increasingly adopted in real-world applications. However, as the zero-shot forecasting paradigm gets popular, a critical yet overlooked question emerges: Are TSFMs robust to adversarial input perturbations? Such perturbations could be exploited in man-in-the-middle attacks or data poisoning. To address this gap, we conduct a systematic investigation into the adversarial robustness of TSFMs. Our results show that even minimal perturbations can induce significant and controllable changes in forecast behaviors, including trend reversal, temporal drift, and amplitude shift, posing serious risks to TSFM-based services. Through experiments on representative TSFMs and multiple datasets, we reveal their consistent vulnerabilities and identify potential architectural designs, such as structural sparsity and multi-task pretraining, that may improve robustness. Our findings offer actionable guidance for designing more resilient forecasting systems and provide a critical assessment of the adversarial robustness of TSFMs.

Yujiang Wu, Hongjian Song, Jiawen Zhang et al.

The development of effective machine learning methodologies for enhancing the efficiency and accuracy of clinical systems is crucial. Despite significant research efforts, managing a plethora of diversified clinical tasks and adapting to emerging new tasks remain significant challenges. This paper presents a novel paradigm that employs a pre-trained large language model as a universal clinical multi-task decoder. This approach leverages the flexibility and diversity of language expressions to handle task topic variations and associated arguments. The introduction of a new task simply requires the addition of a new instruction template. We validate this framework across hundreds of tasks, demonstrating its robustness in facilitating multi-task predictions, performing on par with traditional multi-task learning and single-task learning approaches. Moreover, it shows exceptional adaptability to new tasks, with impressive zero-shot performance in some instances and superior data efficiency in few-shot scenarios. This novel approach offers a unified solution to manage a wide array of new and emerging tasks in clinical applications.