Lei Ma

Yuheng Huang, Jiayang Song, Qiang Hu et al.

Performance evaluation plays a crucial role in the development life cycle of large language models (LLMs). It estimates the model's capability, elucidates behavior characteristics, and facilitates the identification of potential issues and limitations, thereby guiding further improvement. Given that LLMs' diverse task-handling abilities stem from large volumes of training data, a comprehensive evaluation also necessitates abundant, well-annotated, and representative test data to assess LLM performance across various downstream tasks. However, the demand for high-quality test data often entails substantial time, computational resources, and manual efforts, sometimes causing the evaluation to be inefficient or impractical. To address these challenges, researchers propose active testing, which estimates the overall performance by selecting a subset of test data. Nevertheless, the existing active testing methods tend to be inefficient, even inapplicable, given the unique new challenges of LLMs (e.g., diverse task types, increased model complexity, and unavailability of training data). To mitigate such limitations and expedite the development cycle of LLMs, in this work, we introduce AcTracer, an active testing framework tailored for LLMs that strategically selects a small subset of test data to achieve a more accurate performance estimation for LLMs. AcTracer utilizes both internal and external information from LLMs to guide the test sampling process, reducing variance through a multi-stage pool-based active selection. Our experiment results demonstrate that AcTracer achieves state-of-the-art performance compared to existing methods across various tasks.

Xuan Xie, Jiayang Song, Yuheng Huang et al.

Large Language Models (LLMs) are widely used in many different domains, but because of their limited interpretability, there are questions about how trustworthy they are in various perspectives, e.g., truthfulness and toxicity. Recent research has started developing testing methods for LLMs, aiming to uncover untrustworthy issues, i.e., defects, before deployment. However, systematic and formalized testing criteria are lacking, which hinders a comprehensive assessment of the extent and adequacy of testing exploration. To mitigate this threat, we propose a set of multi-level testing criteria, LeCov, for LLMs. The criteria consider three crucial LLM internal components, i.e., the attention mechanism, feed-forward neurons, and uncertainty, and contain nine types of testing criteria in total. We apply the criteria in two scenarios: test prioritization and coverage-guided testing. The experiment evaluation, on three models and four datasets, demonstrates the usefulness and effectiveness of LeCov.

Xindi Zheng, Yuwei Wu, Yu Pan et al.

Missing data imputation poses a paramount challenge when dealing with graph data. Prior works typically are based on feature propagation or graph autoencoders to address this issue. However, these methods usually encounter the over-smoothing issue when dealing with missing data, as the graph neural network (GNN) modules are not explicitly designed for handling missing data. This paper proposes a novel framework, called Dual-Path Generative Adversarial Network (DPGAN), that can deal simultaneously with missing data and avoid over-smoothing problems. The crux of our work is that it admits both global and local representations of the input graph signal, which can capture the long-range dependencies. It is realized via our proposed generator, consisting of two key components, i.e., MLPUNet++ and GraphUNet++. Our generator is trained with a designated discriminator via an adversarial process. In particular, to avoid assessing the entire graph as did in the literature, our discriminator focuses on the local subgraph fidelity, thereby boosting the quality of the local imputation. The subgraph size is adjustable, allowing for control over the intensity of adversarial regularization. Comprehensive experiments across various benchmark datasets substantiate that DPGAN consistently rivals, if not outperforms, existing state-of-the-art imputation algorithms. The code is provided at \url{https://github.com/momoxia/DPGAN}.

Weihao Xuan, Rui Yang, Heli Qi et al.

Existing large language model (LLM) evaluation benchmarks primarily focus on English, while current multilingual tasks lack parallel questions that specifically assess cross-linguistic reasoning abilities. This dual limitation makes it challenging to comprehensively assess LLMs' performance in the multilingual setting. To fill this gap, we introduce MMLU-ProX, a comprehensive benchmark covering 29 languages, built on an English benchmark. Each language version consists of 11,829 identical questions, enabling direct cross-linguistic comparisons. Additionally, to meet efficient evaluation needs, we provide a lite version containing 658 questions per language. To ensure the high quality of MMLU-ProX, we employ a rigorous development process that involves multiple powerful LLMs for translation, followed by expert review to ensure accurate expression, consistent terminology, and cultural relevance. Building on this, we systematically evaluate 36 state-of-the-art LLMs, including reasoning-enhanced and multilingual-optimized LLMs. The results reveal significant disparities in the multilingual capabilities of LLMs: While they perform well in high-resource languages, their performance declines markedly in low-resource languages, with gaps of up to 24.3%. Through MMLU-ProX, we aim to advance the development of more inclusive AI systems and promote equitable access to technology across global contexts.