This paper proposes a simple and robust zero-shot voice conversion system with a cycle structure and mel-spectrogram pre-processing. Previous works suffer from information loss and poor synthesis quality due to their reliance on a carefully designed bottleneck structure. Moreover, models relying solely on self-reconstruction loss struggled with reproducing different speakers' voices. To address these issues, we suggested a cycle-consistency loss that considers conversion back and forth between target and source speakers. Additionally, stacked random-shuffled mel-spectrograms and a label smoothing method are utilized during speaker encoder training to extract a time-independent global speaker representation from speech, which is the key to a zero-shot conversion. Our model outperforms existing state-of-the-art results in both subjective and objective evaluations. Furthermore, it facilitates cross-lingual voice conversions and enhances the quality of synthesized speech.

Since its launch in November 2022, ChatGPT has had a transformative effect on education where students are using it to help with homework assignments and teachers are actively employing it in their teaching practices. This includes using ChatGPT as a tool for writing teachers to grade and generate feedback on students' essays. In this study, we evaluated the quality of the feedback generated by ChatGPT regarding the coherence and cohesion of the essays written by English Language Learners (ELLs) students. We selected 50 argumentative essays and generated feedback on coherence and cohesion using the ELLIPSE rubric. During the feedback evaluation, we used a two-step approach: first, each sentence in the feedback was classified into subtypes based on its function (e.g., positive reinforcement, problem statement). Next, we evaluated its accuracy and usability according to these types. Both the analysis of feedback types and the evaluation of accuracy and usability revealed that most feedback sentences were highly abstract and generic, failing to provide concrete suggestions for improvement. The accuracy in detecting major problems, such as repetitive ideas and the inaccurate use of cohesive devices, depended on superficial linguistic features and was often incorrect. In conclusion, ChatGPT, without specific training for the feedback generation task, does not offer effective feedback on ELL students' coherence and cohesion.

With the increasing capabilities of large language models (LLMs), these high-performance models have achieved state-of-the-art results on a wide range of natural language processing (NLP) tasks. However, the models' performance on commonly-used benchmark datasets often fails to accurately reflect their reliability and robustness when applied to real-world noisy data. To address these challenges, we propose a unified robustness evaluation framework based on the slot-filling task to systematically evaluate the dialogue understanding capability of LLMs in diverse input perturbation scenarios. Specifically, we construct a input perturbation evaluation dataset, Noise-LLM, which contains five types of single perturbation and four types of mixed perturbation data. Furthermore, we utilize a multi-level data augmentation method (character, word, and sentence levels) to construct a candidate data pool, and carefully design two ways of automatic task demonstration construction strategies (instance-level and entity-level) with various prompt templates. Our aim is to assess how well various robustness methods of LLMs perform in real-world noisy scenarios. The experiments have demonstrated that the current open-source LLMs generally achieve limited perturbation robustness performance. Based on these experimental observations, we make some forward-looking suggestions to fuel the research in this direction.

Recently, ChatGPT has attracted great attention from both industry and academia due to its surprising abilities in natural language understanding and generation. We are particularly curious about whether it can achieve promising performance on one of the most complex tasks in aspect-based sentiment analysis, i.e., extracting aspect-category-opinion-sentiment quadruples from texts. To this end, in this paper we develop a specialized prompt template that enables ChatGPT to effectively tackle this complex quadruple extraction task. Further, we propose a selection method on few-shot examples to fully exploit the in-context learning ability of ChatGPT and uplift its effectiveness on this complex task. Finally, we provide a comparative evaluation on ChatGPT against existing state-of-the-art quadruple extraction models based on four public datasets and highlight some important findings regarding the capability boundaries of ChatGPT in the quadruple extraction.

Significant advancements have occurred in the application of Large Language Models (LLMs) for various tasks and social simulations. Despite this, their capacities to coordinate within task-oriented social contexts are under-explored. Such capabilities are crucial if LLMs are to effectively mimic human-like social behavior and produce meaningful results. To bridge this gap, we introduce collaborative generative agents, endowing LLM-based Agents with consistent behavior patterns and task-solving abilities. We situate these agents in a simulated job fair environment as a case study to scrutinize their coordination skills. We propose a novel framework that equips collaborative generative agents with human-like reasoning abilities and specialized skills. Our evaluation demonstrates that these agents show promising performance. However, we also uncover limitations that hinder their effectiveness in more complex coordination tasks. Our work provides valuable insights into the role and evolution of LLMs in task-oriented social simulations.

While large language models (LLMs) exhibit remarkable capabilities across a wide range of tasks, they pose potential safety concerns, such as the ``jailbreak'' problem, wherein malicious instructions can manipulate LLMs to exhibit undesirable behavior. Although several preventive measures have been developed to mitigate the potential risks associated with LLMs, they have primarily focused on English data. In this study, we reveal the presence of multilingual jailbreak challenges within LLMs and consider two potential risk scenarios: unintentional and intentional. The unintentional scenario involves users querying LLMs using non-English prompts and inadvertently bypassing the safety mechanisms, while the intentional scenario concerns malicious users combining malicious instructions with multilingual prompts to deliberately attack LLMs. The experimental results reveal that in the unintentional scenario, the rate of unsafe content increases as the availability of languages decreases. Specifically, low-resource languages exhibit three times the likelihood of encountering harmful content compared to high-resource languages, with both ChatGPT and GPT-4. In the intentional scenario, multilingual prompts can exacerbate the negative impact of malicious instructions, with astonishingly high rates of unsafe output: 80.92\% for ChatGPT and 40.71\% for GPT-4. To handle such a challenge in the multilingual context, we propose a novel \textsc{Self-Defense} framework that automatically generates multilingual training data for safety fine-tuning. Experimental results show that ChatGPT fine-tuned with such data can achieve a substantial reduction in unsafe content generation. Data is available at https://github.com/DAMO-NLP-SG/multilingual-safety-for-LLMs. Warning: This paper contains examples with potentially harmful content.

This draft paper presents a workflow for creating User Personas with Large Language Models, using the results of a Thematic Analysis of qualitative interviews. The proposed workflow uses improved prompting and a larger pool of Themes, compared to previous work conducted by the author for the same task. This is possible due to the capabilities of a recently released LLM which allows the processing of 16 thousand tokens (GPT3.5-Turbo-16k) and also due to the possibility to offer a refined prompting for the creation of Personas. The paper offers details of performing Phase 2 and 3 of Thematic Analysis, and then discusses the improved workflow for creating Personas. The paper also offers some reflections on the relationship between the proposed process and existing approaches to Personas such as the data-driven and qualitative Personas. Moreover, the paper offers reflections on the capacity of LLMs to capture user behaviours and personality traits, from the underlying dataset of qualitative interviews used for the analysis.

Large Language Models (LLMs) have shown remarkable success in various tasks, but concerns about their safety and the potential for generating malicious content have emerged. In this paper, we explore the power of In-Context Learning (ICL) in manipulating the alignment ability of LLMs. We find that by providing just few in-context demonstrations without fine-tuning, LLMs can be manipulated to increase or decrease the probability of jailbreaking, i.e. answering malicious prompts. Based on these observations, we propose In-Context Attack (ICA) and In-Context Defense (ICD) methods for jailbreaking and guarding aligned language model purposes. ICA crafts malicious contexts to guide models in generating harmful outputs, while ICD enhances model robustness by demonstrations of rejecting to answer harmful prompts. Our experiments show the effectiveness of ICA and ICD in increasing or reducing the success rate of adversarial jailbreaking attacks. Overall, we shed light on the potential of ICL to influence LLM behavior and provide a new perspective for enhancing the safety and alignment of LLMs.

Watermark algorithms for large language models (LLMs) have achieved extremely high accuracy in detecting text generated by LLMs. Such algorithms typically involve adding extra watermark logits to the LLM's logits at each generation step. However, prior algorithms face a trade-off between attack robustness and security robustness. This is because the watermark logits for a token are determined by a certain number of preceding tokens; a small number leads to low security robustness, while a large number results in insufficient attack robustness. In this work, we propose a semantic invariant watermarking method for LLMs that provides both attack robustness and security robustness. The watermark logits in our work are determined by the semantics of all preceding tokens. Specifically, we utilize another embedding LLM to generate semantic embeddings for all preceding tokens, and then these semantic embeddings are transformed into the watermark logits through our trained watermark model. Subsequent analyses and experiments demonstrated the attack robustness of our method in semantically invariant settings: synonym substitution and text paraphrasing settings. Finally, we also show that our watermark possesses adequate security robustness. As the quality of text generated by large language models (LLMs) continues to improve, it addresses a multitude of practical challenges on one hand, while simultaneously giving rise to a spectrum of new issues on the other. Therefore, the detection and labeling of machine-generated text have become extremely important. Text watermarking techniques for LLMs usually embed specific information during text generation to allow high-accuracy detection of LLM-generated text. The mainstream approach for embedding such information is to add extra watermark logits on top of the logits generated by the LLM. For example, Kirchenbauer et al. (2023) divide the vocabulary into red and green lists and increase the scores for the green tokens as the watermark logits.

This work introduces a robotics platform which embeds a conversational AI agent in an embodied system for natural language understanding and intelligent decision-making for service tasks; integrating task planning and human-like conversation. The agent is derived from a large language model, which has learned from a vast corpus of general knowledge. In addition to generating dialogue, this agent can interface with the physical world by invoking commands on the robot; seamlessly merging communication and behavior. This system is demonstrated in a free-form tour-guide scenario, in an HRI study combining robots with and without conversational AI capabilities. Performance is measured along five dimensions: overall effectiveness, exploration abilities, scrutinization abilities, receptiveness to personification, and adaptability.