Performance evaluation plays a crucial role in the development lifecycle of large language models (LLMs). It estimates the model’s capability, elucidates behavior characteristics, and facilitates the identification of potential issues and limitations, thereby guiding further improvement. Given that LLMs’ diverse task-handling abilities stem from large volumes of training data, a comprehensive evaluation also necessitates abundant, well-annotated, and representative test data to assess LLM performance across various downstream tasks. However, the demand for high-quality test data often entails substantial time, computational resources, and manual efforts, sometimes causing the evaluation to be inefficient or impractical. To address these challenges, researchers propose active testing, which estimates the overall performance by selecting a subset of test data. Nevertheless, the existing active testing methods tend to be inefficient, even inapplicable, given the unique new challenges of LLMs (e.g., diverse task types, increased model complexity, and unavailability of training data). To mitigate such limitations and expedite the development cycle of LLMs, in this work, we introduce AcTracer, an active testing framework tailored for LLMs that strategically selects a small subset of test data to achieve a more accurate performance estimation for LLMs. AcTracer utilizes both internal and external information from LLMs to guide the test sampling process, reducing variance through a multi-stage pool-based active selection. Our experiment results demonstrate that AcTracer achieves state-of-the-art performance compared to existing methods across various tasks.

Multi-Sensor Fusion (MSF) based perception systems have become the foundation supporting many industrial applications and domains, such as self-driving cars, robotic arms, and unmanned aerial vehicles. With the rapid development of data-driven artificial intelligence (AI), the perception capabilities of MSF have been comprehensively enhanced, especially in understanding complex, dynamic external environments. Similar to traditional software, AI-enabled MSF systems also require rigorous testing. However, existing testing methods are still limited to evaluating the frame-level perception capabilities (e.g., object detection in static scenes) of singlesensor systems (e.g., image-based and point cloud-based systems). Given that many safety-critical intelligent systems, such as selfdriving cars, are operated in dynamic environments where perception systems play an important role, there comes an urgent need to assess their dynamic perception capabilities in understanding and responding to external environmental variations in real-time.To bridge this gap, we design and implement DRIVENCE, an automated metamorphic testing tool for testing the dynamic perception capabilities of MSF-based systems. DRIVENCE accounts for various real-world physical constraints to generate realistic multi-modal test sequences by inserting multiple dynamic traffic participants into the background image and point cloud driving sequences. To diversify testing sequences, we incorporate six driving patterns derived from real-world common driving behaviors into the testing process. We conduct experiments with five SOTA MSF-based tracking systems to evaluate DRIVENCE from the perspectives of (1) generated test cases’ realism, (2) fault detection capabilities, and (3) test efficiency. The results show that DRIVENCE can generate realistic and modality-consistent test driving sequences and effectively detect various dynamic perception errors within MSF systems.

Large language models (LLMs) provide effective solutions in various application scenarios, with the support of retrieval-augmented generation (RAG). However, developers face challenges in integrating LLM and RAG into software systems, due to lacking interface specifications, various requirements from software context, and complicated system management. In this paper, we have conducted a comprehensive study of 100 open-source applications that incorporate LLMs with RAG support, and identified 18 defect patterns. Our study reveals that 77 % of these applications contain more than three types of integration defects that degrade software functionality, efficiency, and security. Guided by our study, we propose systematic guidelines for resolving these defects in software life cycle. We also construct an open-source defect library HYDRANGEA [1].

Large Language Models (LLMs) have demonstrated unprecedented capabilities in code generation. However, there remains a limited understanding of code generation errors that LLMs can produce. To bridge the gap, we conducted an in-depth analysis of code generation errors across six representative LLMs on the HumanEval dataset. Specifically, we first employed open coding and thematic analysis to distill a comprehensive taxonomy of code generation errors. We analyzed two dimensions of error characteristics-semantic characteristics and syntactic characteristics. Our analysis revealed that LLMs often made non-trivial, multi-line code generation errors in various locations and with various root causes. We further analyzed the correlation between these errors and task complexity as well as test pass rate. Our findings highlighted several challenges in locating and fixing code generation errors made by LLMs. In the end, we discussed several future directions to address these challenges.

As Large Language Models (LLMs) continue to revolutionize Natural Language Processing (NLP) applications, critical concerns about their trustworthiness persist, particularly in safety and robustness. To address these challenges, we introduce TrustVis, an automated evaluation framework that provides a comprehensive assessment of LLM trustworthiness. A key feature of our framework is its interactive user interface, designed to offer intuitive visualizations of trustworthiness metrics. By integrating well-known perturbation methods like AutoDAN and employing majority voting across various evaluation methods, TrustVis not only provides reliable results but also makes complex evaluation processes accessible to users. Preliminary case studies on models like Vicuna-7b, Llama2-7b, and GPT-3.5 demonstrate the effectiveness of our framework in identifying safety and robustness vulnerabilities, while the interactive interface allows users to explore results in detail, empowering targeted model improvements. Video Link: https://youtu.be/k1TrBqNVg8g

ABSTRACT Large language models (LLMs) have recently achieved significant success across various application domains, garnering substantial attention from different communities. Unfortunately, many faults still exist that LLMs cannot properly predict. Such faults will harm the usability of LLMs in general and could introduce safety issues in reliability‐critical systems such as autonomous driving systems. How to quickly reveal these faults in real‐world datasets that LLMs could face is important but challenging. The major reason is that the ground truth is necessary but the data labeling process is heavy considering the time and human effort. To handle this problem, in the conventional deep learning testing field, test selection methods have been proposed for efficiently evaluating deep learning models by prioritizing faults. However, despite their importance, the usefulness of these methods on LLMs is unclear and underexplored. In this paper, we conduct the first empirical study to investigate the effectiveness of existing test selection methods for LLMs. We focus on classification tasks because most existing test selection methods target this setting and reliably estimating confidence scores for variable‐length outputs in generative tasks remains challenging. Experimental results on four different tasks (including both code tasks and natural language processing tasks) and four LLMs (e.g., LLaMA3 and GPT‐4) demonstrated that simple methods such as Margin perform well on LLMs, but there is still a big room for improvement. Based on the study, we further propose MuCS, a prompt Mutation‐based prediction Confidence Smoothing framework to boost the test selection capability for LLMs specifically on classification tasks. Concretely, multiple prompt mutation techniques have been proposed to help collect diverse outputs for confidence smoothing. The results show that our proposed framework significantly enhances existing methods with test relative coverage improvement by up to 70.53%.

The recent performance leap of Large Language Models (LLMs) opens up new opportunities across numerous industrial applications and domains. However, the potential erroneous behavior (e.g., the generation of misinformation and hallucination) has also raised severe concerns for the trustworthiness of LLMs, especially in safety-, security- and reliability-sensitive industrial scenarios, potentially hindering real-world adoptions. While uncertainty estimation has shown its potential for interpreting the prediction risks made by classic machine learning (ML) models, the unique characteristics of recent LLMs (e.g., adopting self-attention mechanism as its core, very largescale model size, often used in generative contexts) pose new challenges for the behavior analysis of LLMs. Up to the present, little progress has been made to better understand whether and to what extent uncertainty estimation can help characterize the capability boundary of an LLM, to counteract its undesired behavior, which is considered to be of great importance with the potential wide-range applications of LLMs across industry domains. To bridge the gap, in this paper, we initiate an early exploratory study of the risk assessment of LLMs from the lens of uncertainty. In particular, we conduct a large-scale study with as many as twelve uncertainty estimation methods and eight general LLMs on four NLP tasks and seven programming-capable LLMs on two code generation tasks to investigate to what extent uncertainty estimation techniques could help characterize the prediction risks of LLMs. Our findings confirm the potential of uncertainty estimation for revealing LLMs’ uncertain/nonfactual predictions. The insights derived from our study can pave the way for more advanced analysis and research on LLMs, ultimately aiming at enhancing their trustworthiness.

The recent advancements in Generative AI have significantly advanced the field of text-to-image generation. The state-of-the-art text-to-image model, Stable Diffusion, is now capable of synthesizing high-quality images with a strong sense of aesthetics. Crafting text prompts that align with the model’s interpretation and the user’s intent thus becomes crucial. However, prompting remains challenging for novice users due to the complexity of the stable diffusion model and the non-trivial efforts required for iteratively editing and refining the text prompts. To address these challenges, we propose PromptCharm, a mixed-initiative system that facilitates text-to-image creation through multi-modal prompt engineering and refinement. To assist novice users in prompting, PromptCharm first automatically refines and optimizes the user’s initial prompt. Furthermore, PromptCharm supports the user in exploring and selecting different image styles within a large database. To assist users in effectively refining their prompts and images, PromptCharm renders model explanations by visualizing the model’s attention values. If the user notices any unsatisfactory areas in the generated images, they can further refine the images through model attention adjustment or image inpainting within the rich feedback loop of PromptCharm. To evaluate the effectiveness and usability of PromptCharm, we conducted a controlled user study with 12 participants and an exploratory user study with another 12 participants. These two studies show that participants using PromptCharm were able to create images with higher quality and better aligned with the user’s expectations compared with using two variants of PromptCharm that lacked interaction or visualization support.

The continuous evolution of pre-trained speech models has greatly advanced Speech Emotion Recognition (SER). However, current research typically relies on utterance-level emotion labels, inadequately capturing the complexity of emotions within a single utterance. In this paper, we introduce GMP-TL, a novel SER framework that employs gender-augmented multi-scale pseudo-label (GMP) based transfer learning to mitigate this gap. Specifically, GMPTL initially uses the pre-trained HuBERT, implementing multi-task learning and multi-scale k -means clustering to acquire frame-level GMPs. Subsequently, to fully leverage frame-level GMPs and utterance-level emotion labels, a two-stage model fine-tuning approach is presented to further optimize GMP-TL. Experiments on IEMOCAP show that our GMP-TL attains a WAR of 80.0% and an UAR of 82.0%, achieving superior performance compared to state-of-the-art unimodal SER methods while also yielding comparable results to multimodal SER approaches.

Deep learning (DL) libraries have become the key component in developing and deploying DL-based software nowadays. With the growing popularity of applying DL models in both academia and industry across various domains, any bugs inherent in the DL libraries can potentially cause unexpected server outcomes. As such, there is an urgent demand for improving the software quality of DL libraries. Although there are some existing approaches specifically designed for testing DL libraries, their focus is usually limited to one specific domain, such as computer vision (CV). It is still not very clear how the existing approaches perform in detecting bugs of different DL libraries regarding different task domains and to what extent. To bridge this gap, we first conduct an empirical study on four representative and state-of-the-art DL library testing approaches. Our empirical study results reveal that it is hard for existing approaches to generalize to other task domains. We also find that the test inputs generated by these approaches usually lack diversity, with only a few types of bugs. What is worse, the false-positive rate of existing approaches is also high ( up to 58% ). To address these issues, we propose a guided differential fuzzing approach based on generation , namely, Gandalf . To generate testing inputs across diverse task domains effectively, Gandalf adopts the context-free grammar to ensure validity and utilizes a Deep Q-Network to maximize the diversity. Gandalf also includes 15 metamorphic relations to make it possible for the generated test cases to generalize across different DL libraries. Such a design can decrease the false positives because of the semantic difference for different APIs. We evaluate the effectiveness of Gandalf on nine versions of three representative DL libraries, covering 309 operators from computer vision, natural language processing, and automated speech recognition. The evaluation results demonstrate that Gandalf can effectively and efficiently generate diverse test inputs. Meanwhile, Gandalf successfully detects five categories of bugs with only 3.1% false-positive rates. We report all 49 new unique bugs found during the evaluation to the DL libraries’ developers, and most of these bugs have been confirmed. Details about our empirical study and evaluation results are available on our project website. 1

Over the past decade, Artificial Intelligence (AI) has had great success recently and is being used in a wide range of academic and industrial fields. More recently, Large Language Models (LLMs) have made rapid advancements that have propelled AI to a new level, enabling and empowering even more diverse applications and industrial domains with intelligence, particularly in areas like software engineering and natural language processing. Nevertheless, a number of emerging trustworthiness concerns and issues exhibited in LLMs, e.g., robustness and hallucination, have already recently received much attention, without properly solving which the widespread adoption of LLMs could be greatly hindered in practice. The distinctive characteristics of LLMs, such as the self-attention mechanism, extremely large neural network scale, and autoregressive generation usage contexts, differ from classic AI software based on Convolutional Neural Networks and Recurrent Neural Networks and present new challenges for quality analysis. Up to the present, it still lacks universal and systematic analysis techniques for LLMs despite the urgent industrial demand across diverse domains. Towards bridging such a gap, in this paper, we initiate an early exploratory study and propose a universal analysis framework for LLMs, namedLUNA, which is designed to be general and extensible and enables versatile analysis of LLMs from multiple quality perspectives in a human-interpretable manner. In particular, we first leverage the data from desired trustworthiness perspectives to construct an abstract model as an auxiliary analysis asset and proxy, which is empowered by various abstract model construction methods built-inLUNA. To assess the quality of the abstract model, we collect and define a number of evaluation metrics, aiming at both the abstract model level and the semantics level. Then, the semantics, which is the degree of satisfaction of the LLM w.r.t. the trustworthiness perspective, is bound to and enriches the abstract model with semantics, which enables more detailed analysis applications for diverse purposes, e.g., abnormal behavior detection. To better understand the potential usefulness of our analysis frameworkLUNA, we conduct a large-scale evaluation, the results of which demonstrate that 1) the abstract model has the potential to distinguish normal and abnormal behavior in LLM, 2)LUNAis effective for the real-world analysis of LLMs in practice, and the hyperparameter settings influence the performance, 3) different evaluation metrics are in different correlations with the analysis performance. In order to encourage further studies in the quality assurance of LLMs, we made all of the code and more detailed experimental results data available on the supplementary website of this paperhttps://sites.google.com/view/llm-luna.

Machine Learning (ML) has been widely used in Natural Language Processing (NLP) applications. A fundamental assumption in ML is that training data and real-world data should follow a similar distribution. However, a deployed ML model may suffer from out-of-distribution (OOD) issues due to distribution shifts in the real-world data. Though many algorithms have been proposed to detect OOD data from text corpora, there is still a lack of interactive tool support for ML developers. In this work, we propose DeepLens, an interactive system that helps users detect and explore OOD issues in massive text corpora. Users can efficiently explore different OOD types in DeepLens with the help of a text clustering method. Users can also dig into a specific text by inspecting salient words highlighted through neuron activation analysis. In a within-subjects user study with 24 participants, participants using DeepLens were able to find nearly twice more types of OOD issues accurately with 22% more confidence compared with a variant of DeepLens that has no interaction or visualization support.

Recurrent Neural Networks (RNNs) have been widely used in Natural Language Processing (NLP) tasks given its superior performance on processing sequential data. However, it is challenging to interpret and debug RNNs due to the inherent complexity and the lack of transparency of RNNs. While many explainable AI (XAI) techniques have been proposed for RNNs, most of them only support local explanations rather than global explanations. In this paper, we present DeepSeer, an interactive system that provides both global and local explanations of RNN behavior in multiple tightly-coordinated views for model understanding and debugging. The core of DeepSeer is a state abstraction method that bundles semantically similar hidden states in an RNN model and abstracts the model as a finite state machine. Users can explore the global model behavior by inspecting text patterns associated with each state and the transitions between states. Users can also dive into individual predictions by inspecting the state trace and intermediate prediction results of a given input. A between-subjects user study with 28 participants shows that, compared with a popular XAI technique, LIME, participants using DeepSeer made a deeper and more comprehensive assessment of RNN model behavior, identified the root causes of incorrect predictions more accurately, and came up with more actionable plans to improve the model performance.

In the past few years, Transformer has been widely adopted in many domains and applications because of its impressive performance. Vision Transformer (ViT), a successful and well-known variant, attracts considerable attention from both industry and academia thanks to its record-breaking performance in various vision tasks. However, ViT is also highly nonlinear like other classical neural networks and could be easily fooled by both natural and adversarial perturbations. This limitation could pose a threat to the deployment of ViT in the real industrial environment, especially in safety-critical scenarios. How to improve the robustness of ViT is thus an urgent issue that needs to be addressed. Among all kinds of robustness, patch robustness is defined as giving a reliable output when a random patch in the input domain is perturbed. The perturbation could be natural corruption, such as part of the camera lens being blurred. It could also be a distribution shift, such as an object that does not exist in the training data suddenly appearing in the camera. And in the worst case, there could be a malicious adversarial patch attack that aims to fool the prediction of a machine learning model by arbitrarily modifying pixels within a restricted region of an input image. This kind of attack is also called physical attack, as it is believed to be more real than digital attack. Although there has been some work on patch robustness improvement of Convolutional Neural Network, related studies on its counterpart ViT are still at an early stage as ViT is usually much more complex with far more parameters. It is harder to assess and improve its robustness, not to mention to provide a provable guarantee. In this work, we propose PatchCensor, aiming to certify the patch robustness of ViT by applying exhaustive testing. We try to provide a provable guarantee by considering the worst patch attack scenarios. Unlike empirical defenses against adversarial patches that may be adaptively breached, certified robust approaches can provide a certified accuracy against arbitrary attacks under certain conditions. However, existing robustness certifications are mostly based on robust training, which often requires substantial training efforts and the sacrifice of model performance on normal samples. To bridge the gap, PatchCensor seeks to improve the robustness of the whole system by detecting abnormal inputs instead of training a robust model and asking it to give reliable results for every input, which may inevitably compromise accuracy. Specifically, each input is tested by voting over multiple inferences with different mutated attention masks, where at least one inference is guaranteed to exclude the abnormal patch. This can be seen as complete-coverage testing, which could provide a statistical guarantee on inference at the test time. Our comprehensive evaluation demonstrates that PatchCensor is able to achieve high certified accuracy (e.g., 67.1% on ImageNet for 2%-pixel adversarial patches), significantly outperforming state-of-the-art techniques while achieving similar clean accuracy (81.8% on ImageNet). The clean accuracy is the same as vanilla ViT models. Meanwhile, our technique also supports flexible configurations to handle different adversarial patch sizes by simply changing the masking strategy.