Zobrazeno 1 - 10
of 321
pro vyhledávání: '"Wen, Yuxin"'
Recent advancements in vision-language models (VLMs) have expanded their potential for real-world applications, enabling these models to perform complex reasoning on images. In the widely used fully autoregressive transformer-based models like LLaVA,
Externí odkaz:
http://arxiv.org/abs/2410.14072
Recent breakthroughs in diffusion models have exhibited exceptional image-generation capabilities. However, studies show that some outputs are merely replications of training data. Such replications present potential legal challenges for model owners
Externí odkaz:
http://arxiv.org/abs/2407.21720
Autor:
Chen, Jiuhai, Qadri, Rifaa, Wen, Yuxin, Jain, Neel, Kirchenbauer, John, Zhou, Tianyi, Goldstein, Tom
Most public instruction finetuning datasets are relatively small compared to the closed source datasets used to train industry models. To study questions about finetuning at scale, such as curricula and learning rate cooldown schedules, there is a ne
Externí odkaz:
http://arxiv.org/abs/2406.10323
Autor:
Hans, Abhimanyu, Wen, Yuxin, Jain, Neel, Kirchenbauer, John, Kazemi, Hamid, Singhania, Prajwal, Singh, Siddharth, Somepalli, Gowthami, Geiping, Jonas, Bhatele, Abhinav, Goldstein, Tom
Large language models can memorize and repeat their training data, causing privacy and copyright risks. To mitigate memorization, we introduce a subtle modification to the next-token training objective that we call the goldfish loss. During training,
Externí odkaz:
http://arxiv.org/abs/2406.10209
It is commonplace to produce application-specific models by fine-tuning large pre-trained models using a small bespoke dataset. The widespread availability of foundation model checkpoints on the web poses considerable risks, including the vulnerabili
Externí odkaz:
http://arxiv.org/abs/2404.01231
Synthetic image data generation represents a promising avenue for training deep learning models, particularly in the realm of transfer learning, where obtaining real images within a specific domain can be prohibitively expensive due to privacy and in
Externí odkaz:
http://arxiv.org/abs/2403.19866
It has recently been shown that adversarial attacks on large language models (LLMs) can "jailbreak" the model into making harmful statements. In this work, we argue that the spectrum of adversarial attacks on LLMs is much larger than merely jailbreak
Externí odkaz:
http://arxiv.org/abs/2402.14020
Autor:
An, Bang, Ding, Mucong, Rabbani, Tahseen, Agrawal, Aakriti, Xu, Yuancheng, Deng, Chenghao, Zhu, Sicheng, Mohamed, Abdirisak, Wen, Yuxin, Goldstein, Tom, Huang, Furong
In the burgeoning age of generative AI, watermarks act as identifiers of provenance and artificial content. We present WAVES (Watermark Analysis Via Enhanced Stress-testing), a benchmark for assessing image watermark robustness, overcoming the limita
Externí odkaz:
http://arxiv.org/abs/2401.08573
Autor:
Jain, Neel, Chiang, Ping-yeh, Wen, Yuxin, Kirchenbauer, John, Chu, Hong-Min, Somepalli, Gowthami, Bartoldson, Brian R., Kailkhura, Bhavya, Schwarzschild, Avi, Saha, Aniruddha, Goldblum, Micah, Geiping, Jonas, Goldstein, Tom
We show that language model finetuning can be improved, sometimes dramatically, with a simple augmentation. NEFTune adds noise to the embedding vectors during training. Standard finetuning of LLaMA-2-7B using Alpaca achieves 29.79% on AlpacaEval, whi
Externí odkaz:
http://arxiv.org/abs/2310.05914
Autor:
Jain, Neel, Schwarzschild, Avi, Wen, Yuxin, Somepalli, Gowthami, Kirchenbauer, John, Chiang, Ping-yeh, Goldblum, Micah, Saha, Aniruddha, Geiping, Jonas, Goldstein, Tom
As Large Language Models quickly become ubiquitous, it becomes critical to understand their security vulnerabilities. Recent work shows that text optimizers can produce jailbreaking prompts that bypass moderation and alignment. Drawing from the rich
Externí odkaz:
http://arxiv.org/abs/2309.00614