Fengqi Zhu, Rongzhen Wang, Shen Nie, Xiaolu Zhang, Chunwei Wu, Jun Zhou, Yankai Lin, Ji-Rong Wen, Chongxuan Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

The success of large generative models has driven a paradigm shift, leveraging massive multi-source data to enhance model capabilities. However, the interaction among these sources remains theoretically underexplored. This paper provides a first attempt to fill this gap by rigorously analyzing multi-source training in conditional generative modeling, where each condition represents a distinct data source. Specifically, we establish a general distribution estimation error bound in average total variation distance for conditional maximum likelihood estimation based on the bracketing number. Our result shows that when source distributions share certain similarities and the model is expressive enough, multi-source training guarantees a sharper bound than single-source training. We further instantiate the general theory on conditional Gaussian estimation and deep generative models including autoregressive and flexible energy-based models, by characterizing their bracketing numbers. The results highlight that the number of sources and similarity among source distributions improve the advantage of multi-source training. Simulations and real-world experiments validate our theory.

Gradient-based bilevel programming leverages unrolling differentiation (UD) or implicit function theorem (IFT) to solve hyperparameter optimization (HO) problems, and is proven effective and scalable in practice. To understand their generalization behavior, existing works establish upper bounds on the uniform stability of these algorithms, while their tightness is still unclear. To this end, this paper attempts to establish stability lower bounds for UD-based and IFT-based algorithms. A central technical challenge arises from the dependency of each outer-level update on the concurrent stage of inner optimization in bilevel programming. To address this problem, we introduce lower-bounded expansion properties to characterize the instability in update rules which can serve as general tools for lower-bound analysis. These properties guarantee the hyperparameter divergence at the outer level and the Lipschitz constant of inner output at the inner level in the context of HO. Guided by these insights, we construct a quadratic example that yields tight lower bounds for the UD-based algorithm and meaningful bounds for a representative IFT-based algorithm. Our tight result indicates that uniform stability has reached its limit in stability analysis for the UD-based algorithm.

Autoregressively trained transformers have brought a profound revolution to the world, especially with their in-context learning (ICL) ability to address downstream tasks. Recently, several studies suggest that transformers learn a mesa-optimizer during autoregressive (AR) pretraining to implement ICL. Namely, the forward pass of the trained transformer is equivalent to optimizing an inner objective function in-context. However, whether the practical non-convex training dynamics will converge to the ideal mesa-optimizer is still unclear. Towards filling this gap, we investigate the non-convex dynamics of a one-layer linear causal self-attention model autoregressively trained by gradient flow, where the sequences are generated by an AR process $x_{t+1} = W x_t$. First, under a certain condition of data distribution, we prove that an autoregressively trained transformer learns $W$ by implementing one step of gradient descent to minimize an ordinary least squares (OLS) problem in-context. It then applies the learned $\widehat{W}$ for next-token prediction, thereby verifying the mesa-optimization hypothesis. Next, under the same data conditions, we explore the capability limitations of the obtained mesa-optimizer. We show that a stronger assumption related to the moments of data is the sufficient and necessary condition that the learned mesa-optimizer recovers the distribution. Besides, we conduct exploratory analyses beyond the first data condition and prove that generally, the trained transformer will not perform vanilla gradient descent for the OLS problem. Finally, our simulation results verify the theoretical results.

In this paper, the distributed security control problem of connected vehicle systems (CVSs) is investigated under denial of service (DoS) attacks and uncertain dynamics. DoS attacks usually block communication channels, resulting in the vehicle inability to receive data from the neighbors. In severe cases, it will affect the control performance of CVSs and even cause vehicle collision and life threats. In order to keep the vehicle platoon stable when the DoS attacks happen, we introduce a random characteristic to describe the impact of the packet loss behavior caused by them. Dependent on the length of the lost packets, we propose a security platoon control protocol to deal with it. Furthermore, the security platoon control problem of CVSs is transformed into a stable problem of Markov jump systems (MJSs) with uncertain parameters. Next, the Lyapunov function method and linear matrix inequations (LMI) are used to analyze the internal stability and design controller. Finally, several simulation results are presented to illustrate the effectiveness of the proposed method.

Medical archives processing is a very important task in a medical information system. It generally consists of three steps: medical archives recognition, feature extraction and text classification. In this paper, we focus on empowering the medical archives processing with knowledge graphs. We first build a semantic-rich medical knowledge graph. Then, we recognize texts from medical archives using several popular optical character recognition (OCR) engines, and extract keywords from texts using a knowledge graph based feature extraction algorithm. Third, we define a semantic measure based on knowledge graph to evaluate the similarity between medical texts, and perform the text classification task. This measure can value semantic relatedness between medical documents, to enhance the text classification. We use medical archives collected from real hospitals for validation. The results show that our algorithms can significantly outperform typical baselines that employs only term statistics.