Tool-augmented large language models (LLMs) are rapidly being integrated into real-world applications. Due to the lack of benchmarks, the community has yet to fully understand the hallucination issues within these models. To address this challenge, we introduce a comprehensive diagnostic benchmark, ToolBH. Specifically, we assess the LLM’s hallucinations through two perspectives: depth and breadth. In terms of depth, we propose a multi-level diagnostic process, including (1) solvability detection, (2) solution planning, and (3) missing-tool analysis. For breadth, we consider three scenarios based on the characteristics of the toolset: missing necessary tools, potential tools, and limited functionality tools. Furthermore, we developed seven tasks and collected 700 evaluation samples through multiple rounds of manual annotation. The results show the significant challenges presented by the ToolBH benchmark. The current advanced models Gemini-1.5-Pro and GPT-4o only achieve total scores of 45.3 and 37.0, respectively, on a scale of 100. In this benchmark, larger model parameters do not guarantee better performance; the training data and response strategies also play crucial roles in tool-enhanced LLM scenarios. Our diagnostic analysis indicates that the primary reason for model errors lies in assessing task solvability. Additionally, open-weight models suffer from performance drops with verbose replies, whereas proprietary models excel with longer reasoning.

Multiple-choice question answering (MCQA) for machine reading comprehension (MRC) is challenging. It requires a model to select a correct answer from several candidate options related to text passages or dialogue. To select the correct answer, such models must have the ability to understand natural languages, comprehend textual representations, and infer the relationship between candidate options, questions, and passages. Previous models calculated representations between passages and question-option pairs separately, thereby ignoring the effect of other relation-pairs. In this study, we propose a human reading comprehension attention (HRCA) model and a passage-question-option (PQO) matrix-guided HRCA model called HRCA+ to increase accuracy. The HRCA model updates the information learned from the previous relation-pair to the next relation-pair. HRCA+ utilizes the textual information and the interior relationship between every two parts in a passage, a question, and the corresponding candidate options. Our proposed method outperforms other state-of-the-art methods. On the Semeval-2018 Task 11 dataset, our proposed method improved accuracy levels from 95.8% to 97.2%, and on the DREAM dataset, it improved accuracy levels from 90.4% to 91.6% without extra training data, from 91.8% to 92.6% with extra training data.

This paper describes our participation in SemEval 2020 Task 7 on assessment of humor in edited news headlines, which includes two subtasks, estimating the humor of micro-editd news headlines (subtask A) and predicting the more humorous of the two edited headlines (subtask B). To address these tasks, we propose two systems. The first system adopts a regression-based fine-tuned single-sequence bidirectional encoder representations from transformers (BERT) model with easy data augmentation (EDA), called “BERT+EDA”. The second system adopts a hybrid of a regression-based fine-tuned sequence-pair BERT model and a combined Naive Bayes and support vector machine (SVM) model estimated on term frequency–inverse document frequency (TFIDF) features, called “BERT+NB-SVM”. In this case, no additional training datasets were used, and the BERT+NB-SVM model outperformed BERT+EDA. The official root-mean-square deviation (RMSE) score for subtask A is 0.57369 and ranks 31st out of 48, whereas the best RMSE of BERT+NB-SVM is 0.52429, ranking 7th. For subtask B, we simply use a sequence-pair BERT model, the official accuracy of which is 0.53196 and ranks 25th out of 32.