As the use of Large Language Models (LLMs) becomes more widespread, understanding their self-evaluation of confidence in generated responses becomes increasingly important as it is integral to the reliability of the output of these models. We introduce the concept of Confidence-Probability Alignment, that connects an LLM's internal confidence, quantified by token probabilities, to the confidence conveyed in the model's response when explicitly asked about its certainty. Using various datasets and prompting techniques that encourage model introspection, we probe the alignment between models' internal and expressed confidence. These techniques encompass using structured evaluation scales to rate confidence, including answer options when prompting, and eliciting the model's confidence level for outputs it does not recognize as its own. Notably, among the models analyzed, OpenAI's GPT-4 showed the strongest confidence-probability alignment, with an average Spearman's $\hat{\rho}$ of 0.42, across a wide range of tasks. Our work contributes to the ongoing efforts to facilitate risk assessment in the application of LLMs and to further our understanding of model trustworthiness.
相關內容
In many clinical settings, the use of both Computed Tomography (CT) and Magnetic Resonance (MRI) is necessary to pursue a thorough understanding of the patient's anatomy and to plan a suitable therapeutical strategy; this is often the case in MRI-based radiotherapy, where CT is always necessary to prepare the dose delivery, as it provides the essential information about the radiation absorption properties of the tissues. Sometimes, MRI is preferred to contour the target volumes. However, this approach is often not the most efficient, as it is more expensive, time-consuming and, most importantly, stressful for the patients. To overcome this issue, in this work, we analyse the capabilities of different configurations of Deep Learning models to generate synthetic CT scans from MRI, leveraging the power of Generative Adversarial Networks (GANs) and, in particular, the CycleGAN architecture, capable of working in an unsupervised manner and without paired images, which were not available. Several CycleGAN models were trained unsupervised to generate CT scans from different MRI modalities with and without contrast agents. To overcome the problem of not having a ground truth, distribution-based metrics were used to assess the model's performance quantitatively, together with a qualitative evaluation where physicians were asked to differentiate between real and synthetic images to understand how realistic the generated images were. The results show how, depending on the input modalities, the models can have very different performances; however, models with the best quantitative results, according to the distribution-based metrics used, can generate very difficult images to distinguish from the real ones, even for physicians, demonstrating the approach's potential.
Large Language Models (LLMs) have emerged as a new paradigm for embodied reasoning and control, most recently by generating robot policy code that utilizes a custom library of vision and control primitive skills. However, prior arts fix their skills library and steer the LLM with carefully hand-crafted prompt engineering, limiting the agent to a stationary range of addressable tasks. In this work, we introduce LRLL, an LLM-based lifelong learning agent that continuously grows the robot skill library to tackle manipulation tasks of ever-growing complexity. LRLL achieves this with four novel contributions: 1) a soft memory module that allows dynamic storage and retrieval of past experiences to serve as context, 2) a self-guided exploration policy that proposes new tasks in simulation, 3) a skill abstractor that distills recent experiences into new library skills, and 4) a lifelong learning algorithm for enabling human users to bootstrap new skills with minimal online interaction. LRLL continuously transfers knowledge from the memory to the library, building composable, general and interpretable policies, while bypassing gradient-based optimization, thus relieving the learner from catastrophic forgetting. Empirical evaluation in a simulated tabletop environment shows that LRLL outperforms end-to-end and vanilla LLM approaches in the lifelong setup while learning skills that are transferable to the real world. Project material will become available at the webpage //gtziafas.github.io/LRLL_project.
Retrieval-Augmented Generation (RAG) is a technique that enhances the capabilities of large language models (LLMs) by incorporating external knowledge sources. This method addresses common LLM limitations, including outdated information and the tendency to produce inaccurate "hallucinated" content. However, the evaluation of RAG systems is challenging, as existing benchmarks are limited in scope and diversity. Most of the current benchmarks predominantly assess question-answering applications, overlooking the broader spectrum of situations where RAG could prove advantageous. Moreover, they only evaluate the performance of the LLM component of the RAG pipeline in the experiments, and neglect the influence of the retrieval component and the external knowledge database. To address these issues, this paper constructs a large-scale and more comprehensive benchmark, and evaluates all the components of RAG systems in various RAG application scenarios. Specifically, we have categorized the range of RAG applications into four distinct types-Create, Read, Update, and Delete (CRUD), each representing a unique use case. "Create" refers to scenarios requiring the generation of original, varied content. "Read" involves responding to intricate questions in knowledge-intensive situations. "Update" focuses on revising and rectifying inaccuracies or inconsistencies in pre-existing texts. "Delete" pertains to the task of summarizing extensive texts into more concise forms. For each of these CRUD categories, we have developed comprehensive datasets to evaluate the performance of RAG systems. We also analyze the effects of various components of the RAG system, such as the retriever, the context length, the knowledge base construction, and the LLM. Finally, we provide useful insights for optimizing the RAG technology for different scenarios.
Despite the importance of closely monitoring patients in the Intensive Care Unit (ICU), many aspects are still assessed in a limited manner due to the time constraints imposed on healthcare providers. For example, although excessive visitations during rest hours can potentially exacerbate the risk of circadian rhythm disruption and delirium, it is not captured in the ICU. Likewise, while mobility can be an important indicator of recovery or deterioration in ICU patients, it is only captured sporadically or not captured at all. In the past few years, the computer vision field has found application in many domains by reducing the human burden. Using computer vision systems in the ICU can also potentially enable non-existing assessments or enhance the frequency and accuracy of existing assessments while reducing the staff workload. In this study, we leverage a state-of-the-art noninvasive computer vision system based on depth imaging to characterize ICU visitations and patients' mobility. We then examine the relationship between visitation and several patient outcomes, such as pain, acuity, and delirium. We found an association between deteriorating patient acuity and the incidence of delirium with increased visitations. In contrast, self-reported pain, reported using the Defense and Veteran Pain Rating Scale (DVPRS), was correlated with decreased visitations. Our findings highlight the feasibility and potential of using noninvasive autonomous systems to monitor ICU patients.
Large Language Models (LLMs) have demonstrated their capabilities across various tasks, from language translation to complex reasoning. Understanding and predicting human behavior and biases are crucial for artificial intelligence (AI) assisted systems to provide useful assistance, yet it remains an open question whether these models can achieve this. This paper addresses this gap by leveraging the reasoning and generative capabilities of the LLMs to predict human behavior in two sequential decision-making tasks. These tasks involve balancing between exploitative and exploratory actions and handling delayed feedback, both essential for simulating real-life decision processes. We compare the performance of LLMs with a cognitive instance-based learning (IBL) model, which imitates human experiential decision-making. Our findings indicate that LLMs excel at rapidly incorporating feedback to enhance prediction accuracy. In contrast, the cognitive IBL model better accounts for human exploratory behaviors and effectively captures loss aversion bias, i.e., the tendency to choose a sub-optimal goal with fewer step-cost penalties rather than exploring to find the optimal choice, even with limited experience. The results highlight the benefits of integrating LLMs with cognitive architectures, suggesting that this synergy could enhance the modeling and understanding of complex human decision-making patterns.
The study evaluates the efficacy of Conversational Artificial Intelligence (CAI) in rectifying cognitive biases and recognizing affect in human-AI interactions, which is crucial for digital mental health interventions. Cognitive biases (systematic deviations from normative thinking) affect mental health, intensifying conditions like depression and anxiety. Therapeutic chatbots can make cognitive-behavioral therapy (CBT) more accessible and affordable, offering scalable and immediate support. The research employs a structured methodology with clinical-based virtual case scenarios simulating typical user-bot interactions. Performance and affect recognition were assessed across two categories of cognitive biases: theory of mind biases (anthropomorphization of AI, overtrust in AI, attribution to AI) and autonomy biases (illusion of control, fundamental attribution error, just-world hypothesis). A qualitative feedback mechanism was used with an ordinal scale to quantify responses based on accuracy, therapeutic quality, and adherence to CBT principles. Therapeutic bots (Wysa, Youper) and general-use LLMs (GTP 3.5, GTP 4, Gemini Pro) were evaluated through scripted interactions, double-reviewed by cognitive scientists and a clinical psychologist. Statistical analysis showed therapeutic bots were consistently outperformed by non-therapeutic bots in bias rectification and in 4 out of 6 biases in affect recognition. The data suggests that non-therapeutic chatbots are more effective in addressing some cognitive biases.
This study applies Natural Language Processing techniques, including Latent Dirichlet Allocation, to analyse anonymised maternity incident investigation reports from the Healthcare Safety Investigation Branch. The reports underwent preprocessing, annotation using the Safety Intelligence Research taxonomy, and topic modelling to uncover prevalent topics and detect differences in maternity care across ethnic groups. A combination of offline and online methods was utilised to ensure data protection whilst enabling advanced analysis, with offline processing for sensitive data and online processing for non-sensitive data using the `Claude 3 Opus' language model. Interactive topic analysis and semantic network visualisation were employed to extract and display thematic topics and visualise semantic relationships among keywords. The analysis revealed disparities in care among different ethnic groups, with distinct focus areas for the Black, Asian, and White British ethnic groups. The study demonstrates the effectiveness of topic modelling and NLP techniques in analysing maternity incident investigation reports and highlighting disparities in care. The findings emphasise the crucial role of advanced data analysis in improving maternity care quality and equity.
Purpose:Generative Artificial Intelligence (GAI) models, such as ChatGPT, may inherit or amplify societal biases due to their training on extensive datasets. With the increasing usage of GAI by students, faculty, and staff in higher education institutions (HEIs), it is urgent to examine the ethical issues and potential biases associated with these technologies. Design/Approach/Methods:This scoping review aims to elucidate how biases related to GAI in HEIs have been researched and discussed in recent academic publications. We categorized the potential societal biases that GAI might cause in the field of higher education. Our review includes articles written in English, Chinese, and Japanese across four main databases, focusing on GAI usage in higher education and bias. Findings:Our findings reveal that while there is meaningful scholarly discussion around bias and discrimination concerning LLMs in the AI field, most articles addressing higher education approach the issue superficially. Few articles identify specific types of bias under different circumstances, and there is a notable lack of empirical research. Most papers in our review focus primarily on educational and research fields related to medicine and engineering, with some addressing English education. However, there is almost no discussion regarding the humanities and social sciences. Additionally, a significant portion of the current discourse is in English and primarily addresses English-speaking contexts. Originality/Value:To the best of our knowledge, our study is the first to summarize the potential societal biases in higher education. This review highlights the need for more in-depth studies and empirical work to understand the specific biases that GAI might introduce or amplify in educational settings, guiding the development of more ethical AI applications in higher education.
The Wizard of Oz (WoZ) method is a widely adopted research approach where a human Wizard ``role-plays'' a not readily available technology and interacts with participants to elicit user behaviors and probe the design space. With the growing ability for modern large language models (LLMs) to role-play, one can apply LLMs as Wizards in WoZ experiments with better scalability and lower cost than the traditional approach. However, methodological guidance on responsibly applying LLMs in WoZ experiments and a systematic evaluation of LLMs' role-playing ability are lacking. Through two LLM-powered WoZ studies, we take the first step towards identifying an experiment lifecycle for researchers to safely integrate LLMs into WoZ experiments and interpret data generated from settings that involve Wizards role-played by LLMs. We also contribute a heuristic-based evaluation framework that allows the estimation of LLMs' role-playing ability in WoZ experiments and reveals LLMs' behavior patterns at scale.
Deep Learning algorithms have achieved the state-of-the-art performance for Image Classification and have been used even in security-critical applications, such as biometric recognition systems and self-driving cars. However, recent works have shown those algorithms, which can even surpass the human capabilities, are vulnerable to adversarial examples. In Computer Vision, adversarial examples are images containing subtle perturbations generated by malicious optimization algorithms in order to fool classifiers. As an attempt to mitigate these vulnerabilities, numerous countermeasures have been constantly proposed in literature. Nevertheless, devising an efficient defense mechanism has proven to be a difficult task, since many approaches have already shown to be ineffective to adaptive attackers. Thus, this self-containing paper aims to provide all readerships with a review of the latest research progress on Adversarial Machine Learning in Image Classification, however with a defender's perspective. Here, novel taxonomies for categorizing adversarial attacks and defenses are introduced and discussions about the existence of adversarial examples are provided. Further, in contrast to exisiting surveys, it is also given relevant guidance that should be taken into consideration by researchers when devising and evaluating defenses. Finally, based on the reviewed literature, it is discussed some promising paths for future research.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191