llm - 2024_01
Navigation
- Part 1
- Part 2
Papers
Large Language Models (LLMs) have revolutionized Natural Language Processing but exhibit limitations, particularly in autonomously addressing novel challenges such as reasoning and problem-solving. Traditional techniques like chain-of-thought prompting necessitate explicit human guidance. This paper introduces a novel multi-agent communication framework, inspired by the CAMEL model, to enhance LLMs' autonomous problem-solving capabilities. The framework employs multiple LLM agents, each with a distinct persona, engaged in role-playing communication, offering a nuanced and adaptable approach to diverse problem scenarios. Extensive experimentation demonstrates the framework's superior performance and adaptability, providing valuable insights into the collaborative potential of multiple agents in overcoming the limitations of individual models.
Modern search engines are built on a stack of different components, including query understanding, retrieval, multi-stage ranking, and question answering, among others. These components are often optimized and deployed independently. In this paper, we introduce a novel conceptual framework called large search model, which redefines the conventional search stack by unifying search tasks with one large language model (LLM). All tasks are formulated as autoregressive text generation problems, allowing for the customization of tasks through the use of natural language prompts. This proposed framework capitalizes on the strong language understanding and reasoning capabilities of LLMs, offering the potential to enhance search result quality while simultaneously simplifying the existing cumbersome search stack. To substantiate the feasibility of this framework, we present a series of proof-of-concept experiments and discuss the potential challenges associated with implementing this approach within real-world search systems.
The emergence of LLM (Large Language Model) integrated virtual assistants has brought about a rapid transformation in communication dynamics. During virtual assistant development, some developers prefer to leverage the system message, also known as an initial prompt or custom prompt, for preconditioning purposes. However, it is important to recognize that an excessive reliance on this functionality raises the risk of manipulation by malicious actors who can exploit it with carefully crafted prompts. Such malicious manipulation poses a significant threat, potentially compromising the accuracy and reliability of the virtual assistant's responses. Consequently, safeguarding the virtual assistants with detection and defense mechanisms becomes of paramount importance to ensure their safety and integrity. In this study, we explored three detection and defense mechanisms aimed at countering attacks that target the system message. These mechanisms include inserting a reference key, utilizing an LLM evaluator, and implementing a Self-Reminder. To showcase the efficacy of these mechanisms, they were tested against prominent attack techniques. Our findings demonstrate that the investigated mechanisms are capable of accurately identifying and counteracting the attacks. The effectiveness of these mechanisms underscores their potential in safeguarding the integrity and reliability of virtual assistants, reinforcing the importance of their implementation in real-world scenarios. By prioritizing the security of virtual assistants, organizations can maintain user trust, preserve the integrity of the application, and uphold the high standards expected in this era of transformative technologies.
As large language models (LLM) evolve in their capabilities, various recent studies have tried to quantify their behavior using psychological tools created to study human behavior. One such example is the measurement of "personality" of LLMs using self-assessment personality tests developed to measure human personality. Yet almost none of these works verify the applicability of these tests on LLMs. In this paper, we analyze the reliability of LLM personality scores obtained from self-assessment personality tests using two simple experiments. We first introduce the property of prompt sensitivity, where three semantically equivalent prompts representing three intuitive ways of administering self-assessment tests on LLMs are used to measure the personality of the same LLM. We find that all three prompts lead to very different personality scores, a difference that is statistically significant for all traits in a large majority of scenarios. We then introduce the property of option-order symmetry for personality measurement of LLMs. Since most of the self-assessment tests exist in the form of multiple choice question (MCQ) questions, we argue that the scores should also be robust to not just the prompt template but also the order in which the options are presented. This test unsurprisingly reveals that the self-assessment test scores are not robust to the order of the options. These simple tests, done on ChatGPT and three Llama2 models of different sizes, show that self-assessment personality tests created for humans are unreliable measures of personality in LLMs.