Aligning language models with human preferences is critical for real-world deployment, but existing methods often require large amounts of high-quality human annotations. Aiming at a data-efficient alignment method, we propose Stackelberg Game Preference Optimization (SGPO), a framework that models alignment as a two-player Stackelberg game, where a policy (leader) optimizes against a worst-case preference distribution (follower) within an $\epsilon$-Wasserstein ball, ensuring robustness to (self-)annotation noise and distribution shifts. SGPO guarantees $O(\epsilon)$-bounded regret, unlike Direct Preference Optimization (DPO), which suffers from linear regret growth in the distribution mismatch. We instantiate SGPO with the Stackelberg Self-Annotated Preference Optimization (SSAPO) algorithm, which iteratively self-annotates preferences and adversarially reweights synthetic annotated preferences. Using only 2K seed preferences, from the UltraFeedback dataset, i.e., 1/30 of human labels in the dataset, our method achieves 35.82% GPT-4 win-rate with Mistral-7B and 40.12% with Llama3-8B-Instruct within three rounds of SSAPO.

Prior studies on Visual Sentiment Understanding (VSU) primarily rely on the explicit scene information (e.g., facial expression) to judge visual sentiments, which largely ignore implicit scene information (e.g., human action, objection relation and visual background), while such information is critical for precisely discovering visual sentiments. Motivated by this, this paper proposes a new Omni-scene driven visual Sentiment Identifying, Locating and Attributing in videos (Omni-SILA) task, aiming to interactively and precisely identify, locate and attribute visual sentiments through both explicit and implicit scene information. Furthermore, this paper believes that this Omni-SILA task faces two key challenges: modeling scene and highlighting implicit scene beyond explicit. To this end, this paper proposes an Implicit-enhanced Causal MoE (ICM) approach for addressing the Omni-SILA task. Specifically, a Scene-Balanced MoE (SBM) and an Implicit-Enhanced Causal (IEC) blocks are tailored to model scene information and highlight the implicit scene information beyond explicit, respectively. Extensive experimental results on our constructed explicit and implicit Omni-SILA datasets demonstrate the great advantage of the proposed ICM approach over advanced Video-LLMs.

Electronic health record (EHR) data has emerged as a valuable resource for analyzing patient health status. However, the prevalence of missing data in EHR poses significant challenges to existing methods, leading to spurious correlations and suboptimal predictions. While various imputation techniques have been developed to address this issue, they often obsess unnecessary details and may introduce additional noise when making clinical predictions. To tackle this problem, we propose SMART, a Self-Supervised Missing-Aware RepresenTation Learning approach for patient health status prediction, which encodes missing information via elaborated attentions and learns to impute missing values through a novel self-supervised pre-training approach that reconstructs missing data representations in the latent space. By adopting missing-aware attentions and focusing on learning higher-order representations, SMART promotes better generalization and robustness to missing data.

Parameter-efficient fine-tuning methods, represented by LoRA, play an essential role in adapting large-scale pre-trained models to downstream tasks. However, fine-tuning LoRA-series models also faces the risk of overfitting on the training dataset, and yet there's still a lack of theoretical guidance and practical mechanism to control overfitting on LoRA-based PEFT methods. In this paper, we propose a LoRA Dropout mechanism for the LoRA-based methods by introducing random noises to the learnable low-rank matrices and increasing parameter sparsity. We then demonstrate the theoretical mechanism of our LoRA Dropout mechanism from the perspective of sparsity regularization by providing a generalization error bound under this framework. Theoretical results show that appropriate sparsity would help tighten the gap between empirical and generalization risks and thereby control overfitting. Furthermore, based on the LoRA Dropout framework, we introduce a test-time ensemble strategy and provide theoretical evidence demonstrating that the ensemble method can further compress the error bound, and lead to better performance during inference time. Extensive experiments on various NLP tasks provide practical validations of the effectiveness of our LoRA Dropout framework in improving model accuracy and calibration.