In response to rising concerns surrounding the safety, security, and trustworthiness of Generative AI (GenAI) models, practitioners and regulators alike have pointed to AI red-teaming as a key component of their strategies for identifying and mitigating these risks. However, despite AI red-teaming's central role in policy discussions and corporate messaging, significant questions remain about what precisely it means, what role it can play in regulation, and how it relates to conventional red-teaming practices as originally conceived in the field of cybersecurity. In this work, we identify recent cases of red-teaming activities in the AI industry and conduct an extensive survey of relevant research literature to characterize the scope, structure, and criteria for AI red-teaming practices. Our analysis reveals that prior methods and practices of AI red-teaming diverge along several axes, including the purpose of the activity (which is often vague), the artifact under evaluation, the setting in which the activity is conducted (e.g., actors, resources, and methods), and the resulting decisions it informs (e.g., reporting, disclosure, and mitigation). In light of our findings, we argue that while red-teaming may be a valuable big-tent idea for characterizing GenAI harm mitigations, and that industry may effectively apply red-teaming and other strategies behind closed doors to safeguard AI, gestures towards red-teaming (based on public definitions) as a panacea for every possible risk verge on security theater. To move toward a more robust toolbox of evaluations for generative AI, we synthesize our recommendations into a question bank meant to guide and scaffold future AI red-teaming practices.
相關內容
The proliferation of robot-assisted minimally invasive surgery highlights the need for advanced training tools such as cost-effective robotic endotrainers. Current surgical robots often lack haptic feedback, which is crucial for providing surgeons with a real-time sense of touch. This absence can impact the surgeon's ability to perform delicate operations effectively. To enhance surgical training and address this deficiency, we have integrated a cost-effective haptic feedback system into a robotic endotrainer. This system incorporates both kinesthetic (force) and tactile feedback, improving the fidelity of surgical simulations and enabling more precise control during operations. Our system incorporates an innovative, cost-effective Force/Torque sensor utilizing optoelectronic technology, specifically designed to accurately detect forces and moments exerted on surgical tools with a 95% accuracy, providing essential kinesthetic feedback. Additionally, we implemented a tactile feedback mechanism that informs the surgeon of the gripping forces between the tool's tip and the tissue. This dual feedback system enhances the fidelity of training simulations and the execution of robotic surgeries, promoting broader adoption and safer practices.
Despite the impressive performance in a variety of complex tasks, modern large language models (LLMs) still have trouble dealing with some math problems that are simple and intuitive for humans, such as addition. While we can easily learn basic rules of addition and apply them to new problems of any length, LLMs struggle to do the same. Instead, they may rely on similar cases seen in the training corpus for help. We define these two different reasoning mechanisms as "rule-based reasoning" and "case-based reasoning". Since rule-based reasoning is essential for acquiring systematic generalization ability, we aim to explore exactly whether transformers use rule-based or case-based reasoning for math problems. Through carefully designed intervention experiments on five math tasks, we confirm that transformers are performing case-based reasoning, no matter whether scratchpad is used, which aligns with the previous observations that transformers use subgraph matching/shortcut learning to reason. To mitigate such problems, we propose a Rule-Following Fine-Tuning (RFFT) technique to teach transformers to perform rule-based reasoning. Specifically, we provide explicit rules in the input and then instruct transformers to recite and follow the rules step by step. Through RFFT, we successfully enable LLMs fine-tuned on 1-5 digit addition to generalize to up to 12-digit addition with over 95% accuracy, which is over 40% higher than scratchpad. The significant improvement demonstrates that teaching LLMs to use rules explicitly helps them learn rule-based reasoning and generalize better in length.
In recent years, the importance of smart contract security has been heightened by the increasing number of attacks against them. To address this issue, a multitude of static application security testing (SAST) tools have been proposed for detecting vulnerabilities in smart contracts. However, objectively comparing these tools to determine their effectiveness remains challenging. Existing studies often fall short due to the taxonomies and benchmarks only covering a coarse and potentially outdated set of vulnerability types, which leads to evaluations that are not entirely comprehensive and may display bias. In this paper, we fill this gap by proposing an up-to-date and fine-grained taxonomy that includes 45 unique vulnerability types for smart contracts. Taking it as a baseline, we develop an extensive benchmark that covers 40 distinct types and includes a diverse range of code characteristics, vulnerability patterns, and application scenarios. Based on them, we evaluated 8 SAST tools using this benchmark, which comprises 788 smart contract files and 10,394 vulnerabilities. Our results reveal that the existing SAST tools fail to detect around 50% of vulnerabilities in our benchmark and suffer from high false positives, with precision not surpassing 10%. We also discover that by combining the results of multiple tools, the false negative rate can be reduced effectively, at the expense of flagging 36.77 percentage points more functions. Nevertheless, many vulnerabilities, especially those beyond Access Control and Reentrancy vulnerabilities, remain undetected. We finally highlight the valuable insights from our study, hoping to provide guidance on tool development, enhancement, evaluation, and selection for developers, researchers, and practitioners.
Humans are prone to cognitive distortions -- biased thinking patterns that lead to exaggerated responses to specific stimuli, albeit in very different contexts. This paper demonstrates that advanced Multimodal Large Language Models (MLLMs) exhibit similar tendencies. While these models are designed to respond queries under safety mechanism, they sometimes reject harmless queries in the presence of certain visual stimuli, disregarding the benign nature of their contexts. As the initial step in investigating this behavior, we identify three types of stimuli that trigger the oversensitivity of existing MLLMs: Exaggerated Risk, Negated Harm, and Counterintuitive Interpretation. To systematically evaluate MLLMs' oversensitivity to these stimuli, we propose the Multimodal OverSenSitivity Benchmark (MOSSBench). This toolkit consists of 300 manually collected benign multimodal queries, cross-verified by third-party reviewers (AMT). Empirical studies using MOSSBench on 20 MLLMs reveal several insights: (1). Oversensitivity is prevalent among SOTA MLLMs, with refusal rates reaching up to 76% for harmless queries. (2). Safer models are more oversensitive: increasing safety may inadvertently raise caution and conservatism in the model's responses. (3). Different types of stimuli tend to cause errors at specific stages -- perception, intent reasoning, and safety judgement -- in the response process of MLLMs. These findings highlight the need for refined safety mechanisms that balance caution with contextually appropriate responses, improving the reliability of MLLMs in real-world applications. We make our project available at //turningpoint-ai.github.io/MOSSBench/.
Pre-training with self-supervised models, such as Hidden-unit BERT (HuBERT) and wav2vec 2.0, has brought significant improvements in automatic speech recognition (ASR). However, these models usually require an expensive computational cost to achieve outstanding performance, slowing down the inference speed. To improve the model efficiency, we introduce an early exit scheme for ASR, namely HuBERT-EE, that allows the model to stop the inference dynamically. In HuBERT-EE, multiple early exit branches are added at the intermediate layers. When the intermediate prediction of the early exit branch is confident, the model stops the inference, and the corresponding result can be returned early. We investigate the proper early exiting criterion and fine-tuning strategy to effectively perform early exiting. Experimental results on the LibriSpeech show that HuBERT-EE can accelerate the inference of the HuBERT while simultaneously balancing the trade-off between the performance and the latency.
The behavior of Large Language Models (LLMs) as artificial social agents is largely unexplored, and we still lack extensive evidence of how these agents react to simple social stimuli. Testing the behavior of AI agents in classic Game Theory experiments provides a promising theoretical framework for evaluating the norms and values of these agents in archetypal social situations. In this work, we investigate the cooperative behavior of Llama2 when playing the Iterated Prisoner's Dilemma against random adversaries displaying various levels of hostility. We introduce a systematic methodology to evaluate an LLM's comprehension of the game's rules and its capability to parse historical gameplay logs for decision-making. We conducted simulations of games lasting for 100 rounds, and analyzed the LLM's decisions in terms of dimensions defined in behavioral economics literature. We find that Llama2 tends not to initiate defection but it adopts a cautious approach towards cooperation, sharply shifting towards a behavior that is both forgiving and non-retaliatory only when the opponent reduces its rate of defection below 30%. In comparison to prior research on human participants, Llama2 exhibits a greater inclination towards cooperative behavior. Our systematic approach to the study of LLMs in game theoretical scenarios is a step towards using these simulations to inform practices of LLM auditing and alignment.
We prove that black-box variational inference (BBVI) with control variates, particularly the sticking-the-landing (STL) estimator, converges at a geometric (traditionally called "linear") rate under perfect variational family specification. In particular, we prove a quadratic bound on the gradient variance of the STL estimator, one which encompasses misspecified variational families. Combined with previous works on the quadratic variance condition, this directly implies convergence of BBVI with the use of projected stochastic gradient descent. For the projection operator, we consider a domain with triangular scale matrices, which the projection onto is computable in $\Theta(d)$ time, where $d$ is the dimensionality of the target posterior. We also improve existing analysis on the regular closed-form entropy gradient estimators, which enables comparison against the STL estimator, providing explicit non-asymptotic complexity guarantees for both.
We introduce TorchOpera, a compound AI system for enhancing the safety and quality of prompts and responses for Large Language Models. TorchOpera ensures that all user prompts are safe, contextually grounded, and effectively processed, while enhancing LLM responses to be relevant and high quality. TorchOpera utilizes the vector database for contextual grounding, rule-based wrappers for flexible modifications, and specialized mechanisms for detecting and adjusting unsafe or incorrect content. We also provide a view of the compound AI system to reduce the computational cost. Extensive experiments show that TorchOpera ensures the safety, reliability, and applicability of LLMs in real-world settings while maintaining the efficiency of LLM responses.
Incorporating artificial intelligence (AI) technology, particularly large language models (LLMs), is becoming increasingly vital for developing immersive and interactive metaverse experiences. GPT, a representative LLM developed by OpenAI, is leading LLM development and gaining attention for its potential in building the metaverse. The article delves into the pros and cons of utilizing GPT for metaverse-based education, entertainment, personalization, and support. Dynamic and personalized experiences are possible with this technology, but there are also legitimate privacy, bias, and ethical issues to consider. This article aims to help readers understand the possible influence of GPT, according to its unique technological advantages, on the metaverse and how it may be used to effectively create a more immersive and engaging virtual environment by evaluating these opportunities and obstacles.
Compared with cheap addition operation, multiplication operation is of much higher computation complexity. The widely-used convolutions in deep neural networks are exactly cross-correlation to measure the similarity between input feature and convolution filters, which involves massive multiplications between float values. In this paper, we present adder networks (AdderNets) to trade these massive multiplications in deep neural networks, especially convolutional neural networks (CNNs), for much cheaper additions to reduce computation costs. In AdderNets, we take the $\ell_1$-norm distance between filters and input feature as the output response. The influence of this new similarity measure on the optimization of neural network have been thoroughly analyzed. To achieve a better performance, we develop a special back-propagation approach for AdderNets by investigating the full-precision gradient. We then propose an adaptive learning rate strategy to enhance the training procedure of AdderNets according to the magnitude of each neuron's gradient. As a result, the proposed AdderNets can achieve 74.9% Top-1 accuracy 91.7% Top-5 accuracy using ResNet-50 on the ImageNet dataset without any multiplication in convolution layer.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191