Yue Dong

Eric Malmi, Yue Dong, Jonathan Mallinson et al. · deepmind, mila

Text-editing models have recently become a prominent alternative to seq2seq models for monolingual text-generation tasks such as grammatical error correction, simplification, and style transfer. These tasks share a common trait - they exhibit a large amount of textual overlap between the source and target texts. Text-editing models take advantage of this observation and learn to generate the output by predicting edit operations applied to the source sequence. In contrast, seq2seq models generate outputs word-by-word from scratch thus making them slow at inference time. Text-editing models provide several benefits over seq2seq models including faster inference speed, higher sample efficiency, and better control and interpretability of the outputs. This tutorial provides a comprehensive overview of text-editing models and current state-of-the-art approaches, and analyzes their pros and cons. We discuss challenges related to productionization and how these models can be used to mitigate hallucination and bias, both pressing challenges in the field of text generation.

Erfan Shayegani, Yue Dong, Nael Abu-Ghazaleh · mila

We introduce new jailbreak attacks on vision language models (VLMs), which use aligned LLMs and are resilient to text-only jailbreak attacks. Specifically, we develop cross-modality attacks on alignment where we pair adversarial images going through the vision encoder with textual prompts to break the alignment of the language model. Our attacks employ a novel compositional strategy that combines an image, adversarially targeted towards toxic embeddings, with generic prompts to accomplish the jailbreak. Thus, the LLM draws the context to answer the generic prompt from the adversarial image. The generation of benign-appearing adversarial images leverages a novel embedding-space-based methodology, operating with no access to the LLM model. Instead, the attacks require access only to the vision encoder and utilize one of our four embedding space targeting strategies. By not requiring access to the LLM, the attacks lower the entry barrier for attackers, particularly when vision encoders such as CLIP are embedded in closed-source LLMs. The attacks achieve a high success rate across different VLMs, highlighting the risk of cross-modality alignment vulnerabilities, and the need for new alignment approaches for multi-modal models.

Erfan Shayegani, Md Abdullah Al Mamun, Yu Fu et al.

Large Language Models (LLMs) are swiftly advancing in architecture and capability, and as they integrate more deeply into complex systems, the urgency to scrutinize their security properties grows. This paper surveys research in the emerging interdisciplinary field of adversarial attacks on LLMs, a subfield of trustworthy ML, combining the perspectives of Natural Language Processing and Security. Prior work has shown that even safety-aligned LLMs (via instruction tuning and reinforcement learning through human feedback) can be susceptible to adversarial attacks, which exploit weaknesses and mislead AI systems, as evidenced by the prevalence of `jailbreak' attacks on models like ChatGPT and Bard. In this survey, we first provide an overview of large language models, describe their safety alignment, and categorize existing research based on various learning structures: textual-only attacks, multi-modal attacks, and additional attack methods specifically targeting complex systems, such as federated learning or multi-agent systems. We also offer comprehensive remarks on works that focus on the fundamental sources of vulnerabilities and potential defenses. To make this field more accessible to newcomers, we present a systematic review of existing works, a structured typology of adversarial attack concepts, and additional resources, including slides for presentations on related topics at the 62nd Annual Meeting of the Association for Computational Linguistics (ACL'24).

G M Shahariar, Jia Chen, Jiachen Li et al.

Recent studies show that text-to-image (T2I) models are vulnerable to adversarial attacks, especially with noun perturbations in text prompts. In this study, we investigate the impact of adversarial attacks on different POS tags within text prompts on the images generated by T2I models. We create a high-quality dataset for realistic POS tag token swapping and perform gradient-based attacks to find adversarial suffixes that mislead T2I models into generating images with altered tokens. Our empirical results show that the attack success rate (ASR) varies significantly among different POS tag categories, with nouns, proper nouns, and adjectives being the easiest to attack. We explore the mechanism behind the steering effect of adversarial suffixes, finding that the number of critical tokens and content fusion vary among POS tags, while features like suffix transferability are consistent across categories. We have made our implementation publicly available at - https://github.com/shahariar-shibli/Adversarial-Attack-on-POS-Tags.

Jianfeng Xiang, Xiaoxue Chen, Sicheng Xu et al.

Recent advancements in 3D generative modeling have significantly improved the generation realism, yet the field is still hampered by existing representations, which struggle to capture assets with complex topologies and detailed appearance. This paper present an approach for learning a structured latent representation from native 3D data to address this challenge. At its core is a new sparse voxel structure called O-Voxel, an omni-voxel representation that encodes both geometry and appearance. O-Voxel can robustly model arbitrary topology, including open, non-manifold, and fully-enclosed surfaces, while capturing comprehensive surface attributes beyond texture color, such as physically-based rendering parameters. Based on O-Voxel, we design a Sparse Compression VAE which provides a high spatial compression rate and a compact latent space. We train large-scale flow-matching models comprising 4B parameters for 3D generation using diverse public 3D asset datasets. Despite their scale, inference remains highly efficient. Meanwhile, the geometry and material quality of our generated assets far exceed those of existing models. We believe our approach offers a significant advancement in 3D generative modeling.

Sajib Kumar Saha Joy, Arman Hassan Mahy, Meherin Sultana et al.

In this study, we investigate gender bias in Bangla pretrained language models, a largely under explored area in low-resource languages. To assess this bias, we applied gender-name swapping techniques to existing datasets, creating four manually annotated, task-specific datasets for sentiment analysis, toxicity detection, hate speech detection, and sarcasm detection. By altering names and gender-specific terms, we ensured these datasets were suitable for detecting and mitigating gender bias. We then proposed a joint loss optimization technique to mitigate gender bias across task-specific pretrained models. Our approach was evaluated against existing bias mitigation methods, with results showing that our technique not only effectively reduces bias but also maintains competitive accuracy compared to other baseline approaches. To promote further research, we have made both our implementation and datasets publicly available https://github.com/sajib-kumar/Gender-Bias-Mitigation-From-Bangla-PLM

Haz Sameen Shahgir, Khondker Salman Sayeed, Abhik Bhattacharjee et al.

The advent of Vision Language Models (VLM) has allowed researchers to investigate the visual understanding of a neural network using natural language. Beyond object classification and detection, VLMs are capable of visual comprehension and common-sense reasoning. This naturally led to the question: How do VLMs respond when the image itself is inherently unreasonable? To this end, we present IllusionVQA: a diverse dataset of challenging optical illusions and hard-to-interpret scenes to test the capability of VLMs in two distinct multiple-choice VQA tasks - comprehension and soft localization. GPT4V, the best performing VLM, achieves 62.99% accuracy (4-shot) on the comprehension task and 49.7% on the localization task (4-shot and Chain-of-Thought). Human evaluation reveals that humans achieve 91.03% and 100% accuracy in comprehension and localization. We discover that In-Context Learning (ICL) and Chain-of-Thought reasoning substantially degrade the performance of Gemini-Pro in the localization task. Tangentially, we discover a potential weakness in the ICL capabilities of VLMs: they fail to locate optical illusions even when the correct answer is in the context window as a few-shot example.

Haz Sameen Shahgir, Xianghao Kong, Greg Ver Steeg et al.

The widespread use of Text-to-Image (T2I) models in content generation requires careful examination of their safety, including their robustness to adversarial attacks. Despite extensive research on adversarial attacks, the reasons for their effectiveness remain underexplored. This paper presents an empirical study on adversarial attacks against T2I models, focusing on analyzing factors associated with attack success rates (ASR). We introduce a new attack objective - entity swapping using adversarial suffixes and two gradient-based attack algorithms. Human and automatic evaluations reveal the asymmetric nature of ASRs on entity swap: for example, it is easier to replace "human" with "robot" in the prompt "a human dancing in the rain." with an adversarial suffix, but the reverse replacement is significantly harder. We further propose probing metrics to establish indicative signals from the model's beliefs to the adversarial ASR. We identify conditions that result in a success probability of 60% for adversarial attacks and others where this likelihood drops below 5%.

Yufei Li, Simin Chen, Yanghong Guo et al.

Large Language Models (LLMs) have been widely employed in programming language analysis to enhance human productivity. Yet, their reliability can be compromised by various code distribution shifts, leading to inconsistent outputs. While probabilistic methods are known to mitigate such impact through uncertainty calibration and estimation, their efficacy in the language domain remains underexplored compared to their application in image-based tasks. In this work, we first introduce a large-scale benchmark dataset, incorporating three realistic patterns of code distribution shifts at varying intensities. Then we thoroughly investigate state-of-the-art probabilistic methods applied to CodeLlama using these shifted code snippets. We observe that these methods generally improve the uncertainty awareness of CodeLlama, with increased calibration quality and higher uncertainty estimation~(UE) precision. However, our study further reveals varied performance dynamics across different criteria (e.g., calibration error vs misclassification detection) and trade-off between efficacy and efficiency, highlighting necessary methodological selection tailored to specific contexts.

Yu Fu, Haz Sameen Shahgir, Hui Liu et al.

Recent advances in long-context models (LCMs), designed to handle extremely long input contexts, primarily focus on utilizing external contextual information, often leaving the influence of large language models' intrinsic knowledge underexplored. In this work, we investigate how this intrinsic knowledge affects content generation and demonstrate that its impact becomes increasingly pronounced as context length extends. Furthermore, we show that the model's ability to utilize intrinsic knowledge, which we call intrinsic retrieval ability, does not improve simultaneously with its ability to leverage contextual knowledge through extrinsic retrieval ability. Moreover, better extrinsic retrieval can interfere with the model's ability to use its own knowledge effectively, limiting its full potential. To bridge this gap, we design a simple yet effective Hybrid Needle-in-a-Haystack test that evaluates models based on their capabilities across both retrieval abilities, rather than solely emphasizing extrinsic retrieval ability. Our experimental results reveal that Qwen-2.5 models significantly outperform Llama-3.1 models, demonstrating superior intrinsic retrieval ability. Moreover, even the more powerful Llama-3.1-70B-Instruct model fails to exhibit better performance under LCM conditions, highlighting the importance of evaluating models from a dual-retrieval perspective.