While various multimodal multi-image evaluation datasets have been emerged, but these datasets are primarily based on English, and there has yet to be a Chinese multi-image dataset. To fill this gap, we introduce RealBench, the first Chinese multimodal multi-image dataset, which contains 9393 samples and 69910 images. RealBench distinguishes itself by incorporating real user-generated content, ensuring high relevance to real-world applications. Additionally, the dataset covers a wide variety of scenes, image resolutions, and image structures, further increasing the difficulty of multi-image understanding. Ultimately, we conduct a comprehensive evaluation of RealBench using 21 multimodal LLMs of different sizes, including closed-source models that support multi-image inputs as well as open-source visual and video models. The experimental results indicate that even the most powerful closed-source models still face challenges when handling multi-image Chinese scenarios. Moreover, there remains a noticeable performance gap of around 71.8% on average between open-source visual/video models and closed-source models. These results show that RealBench provides an important research foundation for further exploring multi-image understanding capabilities in the Chinese context. Our datasets will be publicly available.

Text2SQL is a task that translates natural language into SQL statements. Context-dependent Text2SQL offers a more natural database interaction by simulating dialogues between users and databases, with CoSQL and SparC as representative datasets. Yet, these datasets struggle to accurately replicate real-world situations. To address this, we introduce MultiSQL, which extends them in three key aspects: (1) Diverse SQL Operations. We incorporate diverse SQL types such as Create, Update, and Insert to broaden the scope of SQL operations. (2) Schema-Integrated Context. We integrated query context with database schema dependencies to better depict database complexity. (3) Extended Dialogues. We expand dialogue length to better simulate long conversations and complex interactions. This multi-type, schema-integrated, context-dependent Text2SQL dataset comprises nearly 800 dialogue groups and over 9,000 interaction turns across 166 complex databases, offering a better benchmark for interactive user-database dialogue.Addressing MultiSQL’s challenges, we refined evaluation metrics to better capture diverse SQL types and schema dependencies. We designed a prompt framework that leverages historical data and self-refinement to accurately capture the dependency between text queries and database structures. Experiments with GPT-3.5, GPT-4, and LLaMA2-7B show both the effectiveness of our strategies and the challenges of MultiSQL. The datasets is available at https://github.com/grandchicken/MultiSQL.

Relation triplet extraction is a fundamental task in natural language processing that aims to identify semantic relationships between entities in text. It is particularly challenging in the zero-shot setting, i.e., zero-shot relation triplet extraction (ZeroRTE), where the relation sets between training and test are disjoint. Existing methods deal with this task by integrating relations into prompts, which may lack sufficient understanding of the unseen relations. To address these limitations, this paper presents a novel Two-Agent Game (TAG) approach to deliberate and debate the semantics of unseen relations. TAG consists of two agents, a generator and an extractor. They iteratively interact in three key steps: attempting, criticizing, and rectifying. This enables the agents to fully debate and understand the unseen relations. Experimental results demonstrate consistent improvement over ALBERT-Large, BART, andGPT3.5, without incurring additional inference costs in all cases. Remarkably, our method outperforms strong baselines by a significant margin, achieving an impressive 6%-16% increase in F1 scores, particularly when dealingwith FewRel with five unseen relations.

Multimodal large language models (MLLMs) have attracted increasing attention in the past few years, but they may still generate descriptions that include objects not present in the corresponding images, a phenomenon known as object hallucination. To eliminate hallucinations, existing methods manually annotate paired responses with and without hallucinations, and then employ various alignment algorithms to improve the alignment capability between images and text. However, they not only demand considerable computation resources during the finetuning stage but also require expensive human annotation to construct paired data needed by the alignment algorithms. To address these issues, we propose an efficient fine-grained unlearning framework (EFUF), which performs gradient ascent utilizing three tailored losses to eliminate hallucinations without paired data. Extensive experiments show that our method consistently reduces hallucinations while preserving the generation quality with modest computational overhead. Our code and datasets will be publicly available.