Zobrazeno 1 - 10
of 2 118
pro vyhledávání: '"P. Kotha"'
The rise of "jailbreak" attacks on language models has led to a flurry of defenses aimed at preventing undesirable responses. We critically examine the two stages of the defense pipeline: (i) defining what constitutes unsafe outputs, and (ii) enforci
Externí odkaz:
http://arxiv.org/abs/2403.14725
Adversarial training enhances the robustness of Machine Learning (ML) models against adversarial attacks. However, obtaining labeled training and adversarial training data in network/cybersecurity domains is challenging and costly. Therefore, this le
Externí odkaz:
http://arxiv.org/abs/2403.10461
Autor:
Longpre, Shayne, Kapoor, Sayash, Klyman, Kevin, Ramaswami, Ashwin, Bommasani, Rishi, Blili-Hamelin, Borhane, Huang, Yangsibo, Skowron, Aviya, Yong, Zheng-Xin, Kotha, Suhas, Zeng, Yi, Shi, Weiyan, Yang, Xianjun, Southen, Reid, Robey, Alexander, Chao, Patrick, Yang, Diyi, Jia, Ruoxi, Kang, Daniel, Pentland, Sandy, Narayanan, Arvind, Liang, Percy, Henderson, Peter
Independent evaluation and red teaming are critical for identifying the risks posed by generative AI systems. However, the terms of service and enforcement strategies used by prominent AI companies to deter model misuse have disincentives on good fai
Externí odkaz:
http://arxiv.org/abs/2403.04893
Recent approaches to improving the extraction of text embeddings from autoregressive large language models (LLMs) have largely focused on improvements to data, backbone pretrained language models, or improving task-differentiation via instructions. I
Externí odkaz:
http://arxiv.org/abs/2402.15449
We lack a systematic understanding of the effects of fine-tuning (via methods such as instruction-tuning or reinforcement learning from human feedback), particularly on tasks outside the narrow fine-tuning distribution. In a simplified scenario, we d
Externí odkaz:
http://arxiv.org/abs/2309.10105
Autor:
Senthilkumaran, Revanth Krishna, Prashanth, Mridu, Viswanath, Hrishikesh, Kotha, Sathvika, Tiwari, Kshitij, Bera, Aniket
Mass casualty incidents (MCIs) pose a significant challenge to emergency medical services by overwhelming available resources and personnel. Effective victim assessment is the key to minimizing casualties during such a crisis. We introduce ARTEMIS, a
Externí odkaz:
http://arxiv.org/abs/2309.08865
Color maps are a commonly used visualization technique in which data are mapped to optical properties, e.g., color or opacity. Color maps, however, do not explicitly convey structures (e.g., positions and scale of features) within data. Topology-base
Externí odkaz:
http://arxiv.org/abs/2307.08795
Most work on the formal verification of neural networks has focused on bounding the set of outputs that correspond to a given set of inputs (for example, bounded perturbations of a nominal input). However, many use cases of neural network verificatio
Externí odkaz:
http://arxiv.org/abs/2302.01404
Publikováno v:
Clinics and Practice, Vol 14, Iss 3, Pp 906-914 (2024)
The Revised Cardiac Risk Index (RCRI) and the American College of Surgeons (ACS) National Surgical Quality Improvement Program (NSQIP) preoperative risk assessment tools are the most widely used methods for quantifying the risk of major negative peri
Externí odkaz:
https://doaj.org/article/217c4c59b6f4420285ad7e8b2c173788
Publikováno v:
Natural Hazards and Earth System Sciences, Vol 24, Pp 1795-1834 (2024)
Current practice in strong ground motion modelling for probabilistic seismic hazard analysis (PSHA) requires the identification and calibration of empirical models appropriate to the tectonic regimes within the region of application, along with quant
Externí odkaz:
https://doaj.org/article/0a3db59c2e0643b98fe1a584db48f1d0