Search

Jailbreaking is an emerging adversarial attack that bypasses the safety alignment deployed in off-the-shelf large language models (LLMs). A considerable amount of research exists proposing more effective jailbreak attacks, including the recent Greedy Coordinate Gradient (GCG) attack, jailbreak template-based attacks such as using "Do-Anything-Now" (DAN), and multilingual jailbreak. In contrast, the defensive side has been relatively less explored. This paper proposes a lightweight yet practical defense called SELFDEFEND, which can defend against all existing jailbreak attacks with minimal delay for jailbreak prompts and negligible delay for normal user prompts. Our key insight is that regardless of the kind of jailbreak strategies employed, they eventually need to include a harmful prompt (e.g., "how to make a bomb") in the prompt sent to LLMs, and we found that existing LLMs can effectively recognize such harmful prompts that violate their safety policies. Based on this insight, we design a shadow stack that concurrently checks whether a harmful prompt exists in the user prompt and triggers a checkpoint in the normal stack once a token of "No" or a harmful prompt is output. The latter could also generate an explainable LLM response to adversarial prompts. We demonstrate our idea of SELFDEFEND works in various jailbreak scenarios through manual analysis in GPT-3.5/4. We also list three future directions to further enhance SELFDEFEND., Comment: Fixed the bibliography reference issue in our LLM jailbreak defense vision paper submitted on 24 Feb 2024

As new research on Large Language Models (LLMs) continues, it is difficult to keep up with new research and models. To help researchers synthesize the new research many have written survey papers, but even those have become numerous. In this paper, we develop a method to automatically assign survey papers to a taxonomy. We collect the metadata of 144 LLM survey papers and explore three paradigms to classify papers within the taxonomy. Our work indicates that leveraging graph structure information on co-category graphs can significantly outperform the language models in two paradigms; pre-trained language models' fine-tuning and zero-shot/few-shot classifications using LLMs. We find that our model surpasses an average human recognition level and that fine-tuning LLMs using weak labels generated by a smaller model, such as the GCN in this study, can be more effective than using ground-truth labels, revealing the potential of weak-to-strong generalization in the taxonomy classification task., Comment: TL;DR: We collected metadata about LLM surveys and developed a method for categorizing them into a taxonomy, indicating the superiority of graph representation learning over language models and revealing the efficacy of fine-tuning using weak labels

A major threat to the peer-review systems of computer science conferences is the existence of "collusion rings" between reviewers. In such collusion rings, reviewers who have also submitted their own papers to the conference work together to manipulate the conference's paper assignment, with the aim of being assigned to review each other's papers. The most straightforward way that colluding reviewers can manipulate the paper assignment is by indicating their interest in each other's papers through strategic paper bidding. One potential approach to solve this important problem would be to detect the colluding reviewers from their manipulated bids, after which the conference can take appropriate action. While prior work has developed effective techniques to detect other kinds of fraud, no research has yet established that detecting collusion rings is even possible. In this work, we tackle the question of whether it is feasible to detect collusion rings from the paper bidding. To answer this question, we conduct empirical analysis of two realistic conference bidding datasets, including evaluations of existing algorithms for fraud detection in other applications. We find that collusion rings can achieve considerable success at manipulating the paper assignment while remaining hidden from detection: for example, in one dataset, undetected colluders are able to achieve assignment to up to 30% of the papers authored by other colluders. In addition, when 10 colluders bid on all of each other's papers, no detection algorithm outputs a group of reviewers with more than 31% overlap with the true colluders. These results suggest that collusion cannot be effectively detected from the bidding using popular existing tools, demonstrating the need to develop more complex detection algorithms as well as those that leverage additional metadata (e.g., reviewer-paper text-similarity scores).

Recent advancements in large foundation models have remarkably enhanced our understanding of sensory information in open-world environments. In leveraging the power of foundation models, it is crucial for AI research to pivot away from excessive reductionism and toward an emphasis on systems that function as cohesive wholes. Specifically, we emphasize developing Agent AI -- an embodied system that integrates large foundation models into agent actions. The emerging field of Agent AI spans a wide range of existing embodied and agent-based multimodal interactions, including robotics, gaming, and healthcare systems, etc. In this paper, we propose a novel large action model to achieve embodied intelligent behavior, the Agent Foundation Model. On top of this idea, we discuss how agent AI exhibits remarkable capabilities across a variety of domains and tasks, challenging our understanding of learning and cognition. Furthermore, we discuss the potential of Agent AI from an interdisciplinary perspective, underscoring AI cognition and consciousness within scientific discourse. We believe that those discussions serve as a basis for future research directions and encourage broader societal engagement., Comment: 22 pages, 4 figures. arXiv admin note: substantial text overlap with arXiv:2401.03568

Humans have a tendency to see 'human'-like qualities in objects around them. We name our cars, and talk to pets and even household appliances, as if they could understand us as other humans do. This behavior, called anthropomorphism, is also seeing traction in Machine Learning (ML), where human-like intelligence is claimed to be perceived in Large Language Models (LLMs). In this position paper, considering professional incentives, human biases, and general methodological setups, we discuss how the current search for Artificial General Intelligence (AGI) is a perfect storm for over-attributing human-like qualities to LLMs. In several experiments, we demonstrate that the discovery of human-interpretable patterns in latent spaces should not be a surprising outcome. Also in consideration of common AI portrayal in the media, we call for the academic community to exercise extra caution, and to be extra aware of principles of academic integrity, in interpreting and communicating about AI research outcomes., Comment: 20 pages, 15 figures. Preliminary work. Under review by the International Conference on Machine Learning (ICML)

Time series analysis is essential for comprehending the complexities inherent in various real-world systems and applications. Although large language models (LLMs) have recently made significant strides, the development of artificial general intelligence (AGI) equipped with time series analysis capabilities remains in its nascent phase. Most existing time series models heavily rely on domain knowledge and extensive model tuning, predominantly focusing on prediction tasks. In this paper, we argue that current LLMs have the potential to revolutionize time series analysis, thereby promoting efficient decision-making and advancing towards a more universal form of time series analytical intelligence. Such advancement could unlock a wide range of possibilities, including modality switching and time series question answering. We encourage researchers and practitioners to recognize the potential of LLMs in advancing time series analysis and emphasize the need for trust in these related efforts. Furthermore, we detail the seamless integration of time series analysis with existing LLM technologies and outline promising avenues for future research., Comment: 16 pages, 8 figures, 1 table

Mechanistic interpretability (MI) aims to understand AI models by reverse-engineering the exact algorithms neural networks learn. Most works in MI so far have studied behaviors and capabilities that are trivial and token-aligned. However, most capabilities are not that trivial, which advocates for the study of hidden representations inside these networks as the unit of analysis. We do a literature review, formalize representations for features and behaviors, highlight their importance and evaluation, and perform some basic exploration in the mechanistic interpretability of representations. With discussion and exploratory results, we justify our position that studying representations is an important and under-studied field, and that currently established methods in MI are not sufficient to understand representations, thus pushing for the research community to work toward new frameworks for studying representations.

Reliability of AI systems is a fundamental concern for the successful deployment and widespread adoption of AI technologies. Unfortunately, the escalating complexity and heterogeneity of AI hardware systems make them inevitably and increasingly susceptible to hardware faults (e.g., bit flips) that can potentially corrupt model parameters. Given this challenge, this paper aims to answer a critical question: How likely is a parameter corruption to result in an incorrect model output? To systematically answer this question, we propose a novel quantitative metric, Parameter Vulnerability Factor (PVF), inspired by architectural vulnerability factor (AVF) in computer architecture community, aiming to standardize the quantification of AI model resilience/vulnerability against parameter corruptions. We define a model parameter's PVF as the probability that a corruption in that particular model parameter will result in an incorrect output. Similar to AVF, this statistical concept can be derived from statistically extensive and meaningful fault injection (FI) experiments. In this paper, we present several use cases on applying PVF to three types of tasks/models during inference -- recommendation (DLRM), vision classification (CNN), and text classification (BERT). PVF can provide pivotal insights to AI hardware designers in balancing the tradeoff between fault protection and performance/efficiency such as mapping vulnerable AI parameter components to well-protected hardware modules. PVF metric is applicable to any AI model and has a potential to help unify and standardize AI vulnerability/resilience evaluation practice.

Small object detection in aerial imagery presents significant challenges in computer vision due to the minimal data inherent in small-sized objects and their propensity to be obscured by larger objects and background noise. Traditional methods using transformer-based models often face limitations stemming from the lack of specialized databases, which adversely affect their performance with objects of varying orientations and scales. This underscores the need for more adaptable, lightweight models. In response, this paper introduces two innovative approaches that significantly enhance detection and segmentation capabilities for small aerial objects. Firstly, we explore the use of the SAHI framework on the newly introduced lightweight YOLO v9 architecture, which utilizes Programmable Gradient Information (PGI) to reduce the substantial information loss typically encountered in sequential feature extraction processes. The paper employs the Vision Mamba model, which incorporates position embeddings to facilitate precise location-aware visual understanding, combined with a novel bidirectional State Space Model (SSM) for effective visual context modeling. This State Space Model adeptly harnesses the linear complexity of CNNs and the global receptive field of Transformers, making it particularly effective in remote sensing image classification. Our experimental results demonstrate substantial improvements in detection accuracy and processing efficiency, validating the applicability of these approaches for real-time small object detection across diverse aerial scenarios. This paper also discusses how these methodologies could serve as foundational models for future advancements in aerial object recognition technologies. The source code will be made accessible here., Comment: 7 pages, 5 figures

This paper introduces UQA, a novel dataset for question answering and text comprehension in Urdu, a low-resource language with over 70 million native speakers. UQA is generated by translating the Stanford Question Answering Dataset (SQuAD2.0), a large-scale English QA dataset, using a technique called EATS (Enclose to Anchor, Translate, Seek), which preserves the answer spans in the translated context paragraphs. The paper describes the process of selecting and evaluating the best translation model among two candidates: Google Translator and Seamless M4T. The paper also benchmarks several state-of-the-art multilingual QA models on UQA, including mBERT, XLM-RoBERTa, and mT5, and reports promising results. For XLM-RoBERTa-XL, we have an F1 score of 85.99 and 74.56 EM. UQA is a valuable resource for developing and testing multilingual NLP systems for Urdu and for enhancing the cross-lingual transferability of existing models. Further, the paper demonstrates the effectiveness of EATS for creating high-quality datasets for other languages and domains. The UQA dataset and the code are publicly available at www.github.com/sameearif/UQA.