While recent studies have increasingly emphasized the role of reflection in code repair tasks, existing benchmarks still target the repair generation capability of LLMs, lacking fine-grained evaluation of reflection generation capability. To this end, we propose Code Reffix, a benchmark featuring an automated pipeline with oracle reflections and a dual-task protocol to decouple the evaluation of reflection from repair. Through extensive experiments on 14 LLMs and fine-tuning analysis, we aim to pinpoint performance bottlenecks of code repair, quantify reflection quality, and verify the value of reflection optimization. Evaluations reveal that underperforming reflection capabilities of small-scale LLMs remain a major bottleneck for code repair. By quantifying this gap, Code Reffix provides a critical foundation for optimizing LLMs to achieve superior repair performance.

Multimodal aspect-oriented sentiment classification (MABSC) task has garnered significant attention, which aims to identify the sentiment polarities of aspects by combining both language and vision information. However, the limited multimodal data in this task has become a big gap for the vision-language multimodal fusion. While large-scale vision-language pretrained models have been adapted to multiple tasks, their use for MABSC task is still in a nascent stage. In this work, we present an attempt to use the instruction tuning paradigm to MABSC task and leverage the ability of large vision-language models to alleviate the limitation in the fusion of textual and image modalities. To tackle the problem of potential irrelevance between aspects and images, we propose a plug-and-play selector to autonomously choose the most appropriate instruction from the instruction pool, thereby reducing the impact of irrelevant image noise on the final sentiment classification results. We conduct extensive experiments in various scenarios and our model achieves state-of-the-art performance on benchmark datasets, as well as in few-shot settings.

知识图谱问题生成任务是从给定的知识图谱中生成与其相关的问题。目前,知识图谱问题生成模型主要使用基于RNN或Transformer对知识图谱子图进行编码,但这种方式丢失了显式的图结构化信息,在解码器中忽视了局部信息对节点的重要性。本文提出迭代信息传递图编码器来编码子图,获取子图显式的图结构化信息,此外,我们还使用滑动窗口注意力机制提高RNN解码器,提升子图局部信息对节点的重要度。从WQ和PQ数据集上的实验结果看,我们提出的模型比KTG模型在BLEU4指标上分别高出2.16和15.44,证明了该模型的有效性。

Recently, a few studies have discussed the limitations of datasets collected for the task of detecting hate speech from different viewpoints. We intend to contribute to the conversation by providing a consolidated overview of these issues pertaining to the data that debilitate research in this area. Specifically, we discuss how the varying pre-processing steps and the format for making data publicly available result in highly varying datasets that make an objective comparison between studies difficult and unfair. There is currently no study (to the best of our knowledge) focused on comparing the attributes of existing datasets for hate speech detection, outlining their limitations and recommending approaches for future research. This work intends to fill that gap and become the one-stop shop for information regarding hate speech datasets.