Unsupervised domain adaptive pose estimation is a fundamental yet challenging task due to the need to transfer from labeled synthetic data to unlabeled real data. Nevertheless, the underlying pose semantics, which are governed by spatial structure, remain largely consistent across domains. This observation motivates the use of vision-language models, which provide domain-invariant representations that align well with high-level semantic concepts. Motivated by this, we propose CLIP2Pose, a novel framework that leverages the semantic robustness of frozen CLIP encoders to facilitate cross-domain generalization. We first introduce a semantic-driven prompt mechanism that encodes structural priors, domain-specific appearance, and instance-level context into the image representation. This guides the model to focus on semantically meaningful and structurally relevant features. Next, we propose a semantic modulation module that adaptively refines visual features by conditioning them on prompt-derived embeddings, enhancing alignment between semantics and visual patterns. To further bridge the modality and domain gaps, we design a directional alignment loss that encourages consistent structural reasoning across both vision and language representations. Extensive experiments on domain adaptive human body and hand pose benchmarks show that CLIP2Pose achieves state-of-the-art performance.

Yusong Hu, Runmin Ma, Yue Fan, Jinxin Shi, Zongsheng Cao, Yuhao Zhou, Jiakang Yuan, Shuaiyu Zhang, Shiyang Feng, Xiangchao Yan, Shufei Zhang, Wenlong Zhang, Lei Bai, Bo Zhang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Knowledge tracing assesses students’ mastery and predicts future performance based on historical learning data. Traditional methods primarily rely on predefined static associations between concepts and exercises, which struggle to capture potential causal relationships and dynamic learning patterns, leading to reduced prediction accuracy and limited interpretability. To address these issues, this paper proposes a novel dynamic causal inference framework that integrates Gumbel-Softmax sampling with uncertainty estimation, transforming discrete causal relationships into differentiable continuous weights, and quantifying model uncertainty to enhance robustness against noisy data and improve interpretability. Additionally, inspired by item response theory, the model dynamically adjusts students’ latent states by modeling the interaction between student ability and exercises difficulty. Experimental results on three widely-used benchmarks demonstrate that this method achieves state-of-the-art (SOTA) performance in prediction accuracy while also generating interpretable causal relationship weights that provide insights into knowledge acquisition patterns.

Detecting cognitive biases in large language models (LLMs) is a fascinating task that aims to probe the existing cognitive biases within these models. Current methods for detecting cognitive biases in language models generally suffer from incomplete detection capabilities and a restricted range of detectable bias types. To address this issue, we introduced the ‘MindScope’ dataset, which distinctively integrates static and dynamic elements. The static component comprises 5,170 open-ended questions spanning 72 cognitive bias categories. The dynamic component leverages a rule-based, multi-agent communication framework to facilitate the generation of multi-round dialogues. This framework is flexible and readily adaptable for various psychological experiments involving LLMs. In addition, we introduce a multi-agent detection method applicable to a wide range of detection tasks, which integrates Retrieval-Augmented Generation (RAG), competitive debate, and a reinforcement learning-based decision module. Demonstrating substantial effectiveness, this method has shown to improve detection accuracy by as much as 35.10% compared to GPT-4. Codes and appendix are available at https://github.com/2279072142/MindScope.

Few-shot class-incremental learning (FSCIL) aims to learn novel concepts using limited examples without forgetting. However, most studies focus on solving the catastrophic forgetting problem of FSCIL, while neglecting the poor robustness of these algorithms. For example, introducing just a 10% FGSM attack can result in a decrease of more than 30% in accuracy. To tackle this challenge, we propose the Vital Importance Playback for Robust FSCIL (VIP-FSCIL) method. The effectiveness of this approach stems from its consideration of sample importance to construct a replay exemplar set, as well as the reduction of overfitting risks to adversarial examples through knowledge distillation. Our approach thus enhances the robustness of FSCIL models while preserving their generalization. The experimental results indicate that VIP-FSCIL is able to improve the accuracy of the model in perturbed scenarios by 5-9% on CIFAR100 and CUB200 datasets, compared to other methods.

Hierarchical clustering is an important research branch of cluster analysis that has extensive ranges of practical applications. Meanwhile, it still faces problems such as inaccurate, time-consuming, and difficulty in choosing linkage method. In this paper, we present a new Hierarchical Clustering method based on Local Cores and Sharing concept (HCLCS) which takes a "divide-and-merge" framework by first dividing a data set into several small clusters and then merging them hierarchically. To improve the accuracy, the merging process is further divided into two substeps: (1) pre-connect small clusters that belong very likely to the same category, and (2) merge the pre-connected intermediate clusters and the remaining unconnected small clusters in a classical hierarchical way. Extensive experiments on synthetic and real-world data sets show that HCLCS can achieve better performance than existing methods in dealing with data sets with complex structures and is less time-consuming than two state-of-the-art algorithms (SNN-DPC and RSC).