Zobrazeno 1 - 10
of 295
pro vyhledávání: '"Mittal, Prateek"'
Autor:
Panda, Ashwinee, Isik, Berivan, Qi, Xiangyu, Koyejo, Sanmi, Weissman, Tsachy, Mittal, Prateek
Existing methods for adapting large language models (LLMs) to new tasks are not suited to multi-task adaptation because they modify all the model weights -- causing destructive interference between tasks. The resulting effects, such as catastrophic f
Externí odkaz:
http://arxiv.org/abs/2406.16797
Autor:
Nair, Vineet J., Venkataramanan, Venkatesh, Srivastava, Priyank, Sarker, Partha S., Srivastava, Anurag, Marinovici, Laurentiu D., Zha, Jun, Irwin, Christopher, Mittal, Prateek, Williams, John, Poor, H. Vincent, Annaswamy, Anuradha M.
The electricity grid has evolved from a physical system to a cyber-physical system with digital devices that perform measurement, control, communication, computation, and actuation. The increased penetration of distributed energy resources (DERs) tha
Externí odkaz:
http://arxiv.org/abs/2406.14861
Autor:
Xie, Tinghao, Qi, Xiangyu, Zeng, Yi, Huang, Yangsibo, Sehwag, Udari Madhushani, Huang, Kaixuan, He, Luxi, Wei, Boyi, Li, Dacheng, Sheng, Ying, Jia, Ruoxi, Li, Bo, Li, Kai, Chen, Danqi, Henderson, Peter, Mittal, Prateek
Evaluating aligned large language models' (LLMs) ability to recognize and reject unsafe user requests is crucial for safe, policy-compliant deployments. Existing evaluation efforts, however, face three limitations that we address with SORRY-Bench, ou
Externí odkaz:
http://arxiv.org/abs/2406.14598
Data Shapley provides a principled framework for attributing data's contribution within machine learning contexts. However, existing approaches require re-training models on different data subsets, which is computationally intensive, foreclosing thei
Externí odkaz:
http://arxiv.org/abs/2406.11011
Autor:
Qi, Xiangyu, Panda, Ashwinee, Lyu, Kaifeng, Ma, Xiao, Roy, Subhrajit, Beirami, Ahmad, Mittal, Prateek, Henderson, Peter
The safety alignment of current Large Language Models (LLMs) is vulnerable. Relatively simple attacks, or even benign fine-tuning, can jailbreak aligned models. We argue that many of these vulnerabilities are related to a shared underlying issue: saf
Externí odkaz:
http://arxiv.org/abs/2406.05946
Autor:
Qi, Xiangyu, Huang, Yangsibo, Zeng, Yi, Debenedetti, Edoardo, Geiping, Jonas, He, Luxi, Huang, Kaixuan, Madhushani, Udari, Sehwag, Vikash, Shi, Weijia, Wei, Boyi, Xie, Tinghao, Chen, Danqi, Chen, Pin-Yu, Ding, Jeffrey, Jia, Ruoxi, Ma, Jiaqi, Narayanan, Arvind, Su, Weijie J, Wang, Mengdi, Xiao, Chaowei, Li, Bo, Song, Dawn, Henderson, Peter, Mittal, Prateek
The exposure of security vulnerabilities in safety-aligned language models, e.g., susceptibility to adversarial attacks, has shed light on the intricate interplay between AI safety and AI security. Although the two disciplines now come together under
Externí odkaz:
http://arxiv.org/abs/2405.19524
Retrieval-augmented generation (RAG) has been shown vulnerable to retrieval corruption attacks: an attacker can inject malicious passages into retrieval results to induce inaccurate responses. In this paper, we propose RobustRAG as the first defense
Externí odkaz:
http://arxiv.org/abs/2405.15556
Current research on defending against adversarial examples focuses primarily on achieving robustness against a single attack type such as $\ell_2$ or $\ell_{\infty}$-bounded attacks. However, the space of possible perturbations is much larger and cur
Externí odkaz:
http://arxiv.org/abs/2405.01349
Autor:
Panda, Ashwinee, Choquette-Choo, Christopher A., Zhang, Zhengming, Yang, Yaoqing, Mittal, Prateek
When large language models are trained on private data, it can be a significant privacy risk for them to memorize and regurgitate sensitive information. In this work, we propose a new practical data extraction attack that we call "neural phishing". T
Externí odkaz:
http://arxiv.org/abs/2403.00871
Autor:
Wei, Boyi, Huang, Kaixuan, Huang, Yangsibo, Xie, Tinghao, Qi, Xiangyu, Xia, Mengzhou, Mittal, Prateek, Wang, Mengdi, Henderson, Peter
Large language models (LLMs) show inherent brittleness in their safety mechanisms, as evidenced by their susceptibility to jailbreaking and even non-malicious fine-tuning. This study explores this brittleness of safety alignment by leveraging pruning
Externí odkaz:
http://arxiv.org/abs/2402.05162