As LLM-generated text becomes increasingly prevalent on the internet, often containing hallucinations or biases, detecting such content has emerged as a critical area of research.Recent methods have demonstrated impressive performance in detecting text generated entirely by LLMs.However, in real-world scenarios, users often introduce perturbations to the LLM-generated text, and the robustness of existing detection methods against these perturbations has not been sufficiently explored.This paper empirically investigates this challenge and finds that even minor perturbations can severely degrade the performance of current detection methods. To address this issue, we find that the syntactic tree is minimally affected by disturbances and exhibits distinct differences between human-written and LLM-generated text.Therefore, we propose a detection method based on syntactic trees, which can capture features invariant to perturbations.It demonstrates significantly improved robustness against perturbation on the HC3 and GPT-3.5-mixed datasets.Moreover, it also has the shortest time expenditure.We provide the code and data at https://github.com/thulx18/PRDetect.

With the expansion of the application of Large Language Models (LLMs), concerns about their safety have grown among researchers. Numerous studies have demonstrated the potential risks of LLMs generating harmful content and have proposed various safety assessment benchmarks to evaluate these risks. However, the evaluation questions in current benchmarks, especially for Chinese, are too straightforward, making them easily rejected by target LLMs, and difficult to update with practical relevance due to their lack of correlation with real-world events. This hinders the effective application of these benchmarks in continuous evaluation tasks. To address these limitations, we propose SafetyQuizzer, a question-generation framework designed to evaluate the safety of LLMs more sustainably in the Chinese context. SafetyQuizzer leverages a finetuned LLM and jailbreaking attack templates to generate subtly offensive questions, which reduces the decline rate. Additionally, by utilizing retrieval-augmented generation, SafetyQuizzer incorporates the latest real-world events into evaluation questions, improving the adaptability of the benchmarks. Our experiments demonstrate that evaluation questions generated by SafetyQuizzer significantly reduce the decline rate compared to other benchmarks while maintaining a comparable attack success rate. Our code is available at https://github.com/zhichao-stone/SafetyQuizzer. Warning: this paper contains examples that may be offensive or upsetting.

Multi-turn conversational Question Answering (ConvQA) is a practical task that requires the understanding of conversation history, such as previous QA pairs, the passage context, and current question. It can be applied to a variety of scenarios with human-machine dialogue. The major challenge of this task is to require the model to consider the relevant conversation history while understanding the passage. Existing methods usually simply prepend the history to the current question, or use the complicated mechanism to model the history. This article proposes an impression feature, which use the word-level inter attention mechanism to learn multi-oriented information from conversation history to the input sequence, including attention from history tokens to each token of the input sequence, and history turn inter attention from different history turns to each token of the input sequence, and self-attention within input sequence, where the input sequence contains a current question and a passage. Then a feature selection method is designed to enhance the useful history turns of conversation and weaken the unnecessary information. Finally, we demonstrate the effectiveness of the proposed method on the QuAC dataset, analyze the impact of different feature selection methods, and verify the validity and reliability of the proposed features through visualization and human evaluation.