Large Language Models (LLMs) continue to advance in their capabilities, yet this progress is accompanied by a growing array of safety risks. While significant attention has been dedicated to exploiting weaknesses in LLMs through jailbreaking attacks, there remains a paucity of exploration into defending against these attacks. We point out a pivotal factor contributing to the success of jailbreaks: the inherent conflict between the goals of being helpful and ensuring safety. To counter jailbreaking attacks, we propose to integrate goal prioritization at both training and inference stages. Implementing goal prioritization during inference substantially diminishes the Attack Success Rate (ASR) of jailbreaking attacks, reducing it from 66.4% to 2.0% for ChatGPT and from 68.2% to 19.4% for Vicuna-33B, without compromising general performance. Furthermore, integrating the concept of goal prioritization into the training phase reduces the ASR from 71.0% to 6.6% for LLama2-13B. Remarkably, even in scenarios where no jailbreaking samples are included during training, our approach slashes the ASR by half, decreasing it from 71.0% to 34.0%. Additionally, our findings reveal that while stronger LLMs face greater safety risks, they also possess a greater capacity to be steered towards defending against such attacks. We hope our work could contribute to the comprehension of jailbreaking attacks and defenses, and shed light on the relationship between LLMs' capability and safety. Our code will be available at \url{https://github.com/thu-coai/JailbreakDefense_GoalPriority}.

Large Language Models (LLMs), trained predominantly on extensive English data, often exhibit limitations when applied to other languages. Current research is primarily focused on enhancing the multilingual capabilities of these models by employing various tuning strategies. Despite their effectiveness in certain languages, the understanding of the multilingual abilities of LLMs remains incomplete. This study endeavors to evaluate the multilingual capacity of LLMs by conducting an exhaustive analysis across 101 languages, and classifies languages with similar characteristics into four distinct quadrants. By delving into each quadrant, we shed light on the rationale behind their categorization and offer actionable guidelines for tuning these languages. Extensive experiments reveal that existing LLMs possess multilingual capabilities that surpass our expectations, and we can significantly improve the multilingual performance of LLMs by focusing on these distinct attributes present in each quadrant.

Knowledge Editing (KE) for modifying factual knowledge in Large Language Models (LLMs) has been receiving increasing attention. However, existing knowledge editing methods are entity-centric, and it is unclear whether this approach is suitable for a relation-centric perspective. To address this gap, this paper constructs a new benchmark named RaKE, which focuses on Relation based Knowledge Editing. In this paper, we establish a suite of innovative metrics for evaluation and conduct comprehensive experiments involving various knowledge editing baselines. We notice that existing knowledge editing methods exhibit the potential difficulty in their ability to edit relations. Therefore, we further explore the role of relations in factual triplets within the transformer. Our research results confirm that knowledge related to relations is not only stored in the FFN network but also in the attention layers. This provides experimental support for future relation-based knowledge editing methods.

Recent benchmarks for Large Language Models (LLMs) have mostly focused on application-driven tasks such as complex reasoning and code generation, and this has led to a scarcity in purely linguistic evaluation of LLMs. Against this background, we introduce Multilingual Evaluation of Linguistic Acceptability -- MELA, the first multilingual benchmark on linguistic acceptability with 48K samples covering 10 languages from a diverse set of language families. We establish baselines of commonly used LLMs along with supervised models, and conduct cross-lingual transfer and multi-task learning experiments with XLM-R. In pursuit of multilingual interpretability, we analyze the weights of fine-tuned XLM-R to explore the possibility of identifying transfer difficulty between languages. Our results show that ChatGPT benefits much from in-context examples but still lags behind fine-tuned XLM-R, while the performance of GPT-4 is on par with fine-tuned XLM-R even in zero-shot setting. Cross-lingual and multi-task learning experiments show that unlike semantic tasks, in-language training data is crucial in acceptability judgements. Results in layerwise probing indicate that the upper layers of XLM-R become a task-specific but language-agnostic region for multilingual acceptability judgment. We also introduce the concept of conflicting weight, which could be a potential indicator for the difficulty of cross-lingual transfer between languages. Our data will be available at https://github.com/sjtu-compling/MELA.

Computational argumentation has become an essential tool in various fields, including artificial intelligence, law, and public policy. It is an emerging research field in natural language processing (NLP) that attracts increasing attention. Research on computational argumentation mainly involves two types of tasks: argument mining and argument generation. As large language models (LLMs) have demonstrated strong abilities in understanding context and generating natural language, it is worthwhile to evaluate the performance of LLMs on various computational argumentation tasks. This work aims to embark on an assessment of LLMs, such as ChatGPT, Flan models and LLaMA2 models, under zero-shot and few-shot settings within the realm of computational argumentation. We organize existing tasks into 6 main classes and standardise the format of 14 open-sourced datasets. In addition, we present a new benchmark dataset on counter speech generation, that aims to holistically evaluate the end-to-end performance of LLMs on argument mining and argument generation. Extensive experiments show that LLMs exhibit commendable performance across most of these datasets, demonstrating their capabilities in the field of argumentation. We also highlight the limitations in evaluating computational argumentation and provide suggestions for future research directions in this field.

End-to-end task-oriented dialogue (EToD) can directly generate responses in an end-to-end fashion without modular training, which attracts escalating popularity. The advancement of deep neural networks, especially the successful use of large pre-trained models, has further led to significant progress in EToD research in recent years. In this paper, we present a thorough review and provide a unified perspective to summarize existing approaches as well as recent trends to advance the development of EToD research. The contributions of this paper can be summarized: (1) \textbf{\textit{First survey}}: to our knowledge, we take the first step to present a thorough survey of this research field; (2) \textbf{\textit{New taxonomy}}: we first introduce a unified perspective for EToD, including (i) \textit{Modularly EToD} and (ii) \textit{Fully EToD}; (3) \textbf{\textit{New Frontiers}}: we discuss some potential frontier areas as well as the corresponding challenges, hoping to spur breakthrough research in EToD field; (4) \textbf{\textit{Abundant resources}}: we build a public website\footnote{We collect the related papers, baseline projects, and leaderboards for the community at \url{https://etods.net/}.}, where EToD researchers could directly access the recent progress. We hope this work can serve as a thorough reference for the EToD research community.

Large language models achieve high performance on many but not all downstream tasks. The interaction between pretraining data and task data is commonly assumed to determine this variance: a task with data that is more similar to a model's pretraining data is assumed to be easier for that model. We test whether distributional and example-specific similarity measures (embedding-, token- and model-based) correlate with language model performance through a large-scale comparison of the Pile and C4 pretraining datasets with downstream benchmarks. Similarity correlates with performance for multilingual datasets, but in other benchmarks, we surprisingly find that similarity metrics are not correlated with accuracy or even each other. This suggests that the relationship between pretraining data and downstream tasks is more complex than often assumed.

The increased use of large language models (LLMs) across a variety of real-world applications calls for mechanisms to verify the factual accuracy of their outputs. In this work, we present a holistic end-to-end solution for annotating the factuality of LLM-generated responses, which encompasses a multi-stage annotation scheme designed to yield detailed labels concerning the verifiability and factual inconsistencies found in LLM outputs. We design and build an annotation tool to speed up the labelling procedure and ease the workload of raters. It allows flexible incorporation of automatic results in any stage, e.g. automatically-retrieved evidence. We further construct an open-domain document-level factuality benchmark in three-level granularity: claim, sentence and document. Preliminary experiments show that FacTool, FactScore and Perplexity.ai are struggling to identify false claims with the best F1=0.53. Annotation tool, benchmark and code are available at https://github.com/yuxiaw/Factcheck-GPT.

In-context learning (ICL) has become the default method for using large language models (LLMs), making the exploration of its limitations and understanding the underlying causes crucial. In this paper, we find that ICL falls short of handling specification-heavy tasks, which are tasks with complicated and extensive task specifications, requiring several hours for ordinary humans to master, such as traditional information extraction tasks. The performance of ICL on these tasks mostly cannot reach half of the state-of-the-art results. To explore the reasons behind this failure, we conduct comprehensive experiments on 18 specification-heavy tasks with various LLMs and identify three primary reasons: inability to specifically understand context, misalignment in task schema comprehension with humans, and inadequate long-text understanding ability. Furthermore, we demonstrate that through fine-tuning, LLMs can achieve decent performance on these tasks, indicating that the failure of ICL is not an inherent flaw of LLMs, but rather a drawback of existing alignment methods that renders LLMs incapable of handling complicated specification-heavy tasks via ICL. To substantiate this, we perform dedicated instruction tuning on LLMs for these tasks and observe a notable improvement. We hope the analyses in this paper could facilitate advancements in alignment methods enabling LLMs to meet more sophisticated human demands.

Large language models (LLMs) have shown extraordinary performance in various language tasks, but high computational requirements hinder their widespread deployment. Speculative decoding, which uses amateur models to predict the generation of expert models, has been proposed as a way to accelerate LLM inference. However, speculative decoding focuses on acceleration instead of making the best use of the token distribution from amateur models. We proposed Speculative Contrastive Decoding (SCD), an accelerated decoding method leveraging the natural contrast between expert and amateur models in speculative decoding. Comprehensive evaluations on four benchmarks show that SCD can achieve similar acceleration factors as speculative decoding while further improving the generation quality as the contrastive decoding. The analysis of token probabilities further demonstrates the compatibility between speculative and contrastive decoding. Overall, SCD provides an effective approach to enhance the decoding quality of LLMs while saving computational resources.