Which LLM to Play? Convergence-Aware Online Model Selection with Time-Increasing Bandits

Web-based applications such as chatbots, search engines and news recommendations continue to grow in scale and complexity with the recent surge in the adoption of LLMs. Online model selection has thus garnered increasing attention due to the need to choose the best model among a diverse set while balancing task reward and exploration cost. Organizations faces decisions like whether to employ a costly API-based LLM or a locally finetuned small LLM, weighing cost against performance. Traditional selection methods often evaluate every candidate model before choosing one, which are becoming impractical given the rising costs of training and finetuning LLMs. Moreover, it is undesirable to allocate excessive resources towards exploring poor-performing models. While some recent works leverage online bandit algorithm to manage such exploration-exploitation trade-off in model selection, they tend to overlook the increasing-then-converging trend in model performances as the model is iteratively finetuned, leading to less accurate predictions and suboptimal model selections. In this paper, we propose a time-increasing bandit algorithm TI-UCB, which effectively predicts the increase of model performances due to finetuning and efficiently balances exploration and exploitation in model selection. To further capture the converging points of models, we develop a change detection mechanism by comparing consecutive increase predictions. We theoretically prove that our algorithm achieves a logarithmic regret upper bound in a typical increasing bandit setting, which implies a fast convergence rate. The advantage of our method is also empirically validated through extensive experiments on classification model selection and online selection of LLMs. Our results highlight the importance of utilizing increasing-then-converging pattern for more efficient and economic model selection in the deployment of LLMs.

ChatEL: Entity Linking with Chatbots

Entity Linking (EL) is an essential and challenging task in natural language processing that seeks to link some text representing an entity within a document or sentence with its corresponding entry in a dictionary or knowledge base. Most existing approaches focus on creating elaborate contextual models that look for clues the words surrounding the entity-text to help solve the linking problem. Although these fine-tuned language models tend to work, they can be unwieldy, difficult to train, and do not transfer well to other domains. Fortunately, Large Language Models (LLMs) like GPT provide a highly-advanced solution to the problems inherent in EL models, but simply naive prompts to LLMs do not work well. In the present work, we define ChatEL, which is a three-step framework to prompt LLMs to return accurate results. Overall the ChatEL framework improves the average F1 performance across 10 datasets by more than 2%. Finally, a thorough error analysis shows many instances with the ground truth labels were actually incorrect, and the labels predicted by ChatEL were actually correct. This indicates that the quantitative results presented in this paper may be a conservative estimate of the actual performance. All data and code are available as an open-source package on GitHub at https://github.com/yifding/In_Context_EL.

Citation-Enhanced Generation for LLM-based Chatbot

Large language models (LLMs) exhibit powerful general intelligence across diverse scenarios, including their integration into chatbots. However, a vital challenge of LLM-based chatbots is that they may produce hallucinated content in responses, which significantly limits their applicability. Various efforts have been made to alleviate hallucination, such as retrieval augmented generation and reinforcement learning with human feedback, but most of them require additional training and data annotation. In this paper, we propose a novel post-hoc \textbf{C}itation-\textbf{E}nhanced \textbf{G}eneration (\textbf{CEG}) approach combined with retrieval argumentation. Unlike previous studies that focus on preventing hallucinations during generation, our method addresses this issue in a post-hoc way. It incorporates a retrieval module to search for supporting documents relevant to the generated content, and employs a natural language inference-based citation generation module. Once the statements in the generated content lack of reference, our model can regenerate responses until all statements are supported by citations. Note that our method is a training-free plug-and-play plugin that is capable of various LLMs. Experiments on various hallucination-related datasets show our framework outperforms state-of-the-art methods in both hallucination detection and response regeneration on three benchmarks. Our codes and dataset will be publicly available.

Task-Oriented Dialogue with In-Context Learning

We describe a system for building task-oriented dialogue systems combining the in-context learning abilities of large language models (LLMs) with the deterministic execution of business logic. LLMs are used to translate between the surface form of the conversation and a domain-specific language (DSL) which is used to progress the business logic. We compare our approach to the intent-based NLU approach predominantly used in industry today. Our experiments show that developing chatbots with our system requires significantly less effort than established approaches, that these chatbots can successfully navigate complex dialogues which are extremely challenging for NLU-based systems, and that our system has desirable properties for scaling task-oriented dialogue systems to a large number of tasks. We make our implementation available for use and further study.

Making Them Ask and Answer: Jailbreaking Large Language Models in Few Queries via Disguise and Reconstruction

In recent years, large language models (LLMs) have demonstrated notable success across various tasks, but the trustworthiness of LLMs is still an open problem. One specific threat is the potential to generate toxic or harmful responses. Attackers can craft adversarial prompts that induce harmful responses from LLMs. In this work, we pioneer a theoretical foundation in LLMs security by identifying bias vulnerabilities within the safety fine-tuning and design a black-box jailbreak method named DRA (Disguise and Reconstruction Attack), which conceals harmful instructions through disguise and prompts the model to reconstruct the original harmful instruction within its completion. We evaluate DRA across various open-source and close-source models, showcasing state-of-the-art jailbreak success rates and attack efficiency. Notably, DRA boasts a 90\% attack success rate on LLM chatbots GPT-4.

In Generative AI we Trust: Can Chatbots Effectively Verify Political Information?

This article presents a comparative analysis of the ability of two large language model (LLM)-based chatbots, ChatGPT and Bing Chat, recently rebranded to Microsoft Copilot, to detect veracity of political information. We use AI auditing methodology to investigate how chatbots evaluate true, false, and borderline statements on five topics: COVID-19, Russian aggression against Ukraine, the Holocaust, climate change, and LGBTQ+ related debates. We compare how the chatbots perform in high- and low-resource languages by using prompts in English, Russian, and Ukrainian. Furthermore, we explore the ability of chatbots to evaluate statements according to political communication concepts of disinformation, misinformation, and conspiracy theory, using definition-oriented prompts. We also systematically test how such evaluations are influenced by source bias which we model by attributing specific claims to various political and social actors. The results show high performance of ChatGPT for the baseline veracity evaluation task, with 72 percent of the cases evaluated correctly on average across languages without pre-training. Bing Chat performed worse with a 67 percent accuracy. We observe significant disparities in how chatbots evaluate prompts in high- and low-resource languages and how they adapt their evaluations to political communication concepts with ChatGPT providing more nuanced outputs than Bing Chat. Finally, we find that for some veracity detection-related tasks, the performance of chatbots varied depending on the topic of the statement or the source to which it is attributed. These findings highlight the potential of LLM-based chatbots in tackling different forms of false information in online environments, but also points to the substantial variation in terms of how such potential is realized due to specific factors, such as language of the prompt or the topic.

Exploring consumers response to text-based chatbots in e-commerce: The moderating role of task complexity and chatbot disclosure

Artificial intelligence based chatbots have brought unprecedented business potential. This study aims to explore consumers trust and response to a text-based chatbot in ecommerce, involving the moderating effects of task complexity and chatbot identity disclosure. A survey method with 299 useable responses was conducted in this research. This study adopted the ordinary least squares regression to test the hypotheses. First, the consumers perception of both the empathy and friendliness of the chatbot positively impacts their trust in it. Second, task complexity negatively moderates the relationship between friendliness and consumers trust. Third, disclosure of the text based chatbot negatively moderates the relationship between empathy and consumers trust, while it positively moderates the relationship between friendliness and consumers trust. Fourth, consumers trust in the chatbot increases their reliance on the chatbot and decreases their resistance to the chatbot in future interactions. Adopting the stimulus organism response framework, this study provides important insights on consumers perception and response to the text-based chatbot. The findings of this research also make suggestions that can increase consumers positive responses to text based chatbots. Extant studies have investigated the effects of automated bots attributes on consumers perceptions. However, the boundary conditions of these effects are largely ignored. This research is one of the first attempts to provide a deep understanding of consumers responses to a chatbot.

Measuring and Controlling Persona Drift in Language Model Dialogs

Prompting is a standard tool for customizing language-model chatbots, enabling them to take on a specific "persona". An implicit assumption in the use of prompts is that they will be stable, so the chatbot will continue to generate text according to the stipulated persona for the duration of a conversation. We propose a quantitative benchmark to test this assumption, evaluating persona stability via self-chats between two personalized chatbots. Testing popular models like LLaMA2-chat-70B, we reveal a significant persona drift within eight rounds of conversations. An empirical and theoretical analysis of this phenomenon suggests the transformer attention mechanism plays a role, due to attention decay over long exchanges. To combat attention decay and persona drift, we propose a lightweight method called split-softmax, which compares favorably against two strong baselines.

Comparing Generative Chatbots Based on Process Requirements

Business processes are commonly represented by modelling languages, such as Event-driven Process Chain (EPC), Yet Another Workflow Language (YAWL), and the most popular standard notation for modelling business processes, the Business Process Model and Notation (BPMN). Most recently, chatbots, programs that allow users to interact with a machine using natural language, have been increasingly used for business process execution support. A recent category of chatbots worth mentioning is generative-based chatbots, powered by Large Language Models (LLMs) such as OpenAI's Generative Pre-Trained Transformer (GPT) model and Google's Pathways Language Model (PaLM), which are trained on billions of parameters and support conversational intelligence. However, it is not clear whether generative-based chatbots are able to understand and meet the requirements of constructs such as those provided by BPMN for process execution support. This paper presents a case study to compare the performance of prominent generative models, GPT and PaLM, in the context of process execution support. The research sheds light into the challenging problem of using conversational approaches supported by generative chatbots as a means to understand process-aware modelling notations and support users to execute their tasks.

How Generative-AI can be Effectively used in Government Chatbots

With the rapid development of artificial intelligence and breakthroughs in machine learning and natural language processing, intelligent question-answering robots have become widely used in government affairs. This paper conducts a horizontal comparison between Guangdong Province's government chatbots, ChatGPT, and Wenxin Ernie, two large language models, to analyze the strengths and weaknesses of existing government chatbots and AIGC technology. The study finds significant differences between government chatbots and large language models. China's government chatbots are still in an exploratory stage and have a gap to close to achieve "intelligence." To explore the future direction of government chatbots more deeply, this research proposes targeted optimization paths to help generative AI be effectively applied in government chatbot conversations.