Zobrazeno 1 - 10
of 11 733
pro vyhledávání: '"LEWIS, D A"'
The success of Large Language Models (LLMs) in multicultural environments hinges on their ability to understand users' diverse cultural backgrounds. We measure this capability by having an LLM simulate human profiles representing various nationalitie
Externí odkaz:
http://arxiv.org/abs/2408.06929
Autor:
Griffin, Lewis D, Riggs, Nicholas
Matrix Games are a type of unconstrained wargame used by planners to explore scenarios. Players propose actions, and give arguments and counterarguments for their success. An umpire, assisted by dice rolls modified according to the offered arguments,
Externí odkaz:
http://arxiv.org/abs/2405.10997
While self-supervised learning has improved anomaly detection in computer vision and natural language processing, it is unclear whether tabular data can benefit from it. This paper explores the limitations of self-supervision for tabular anomaly dete
Externí odkaz:
http://arxiv.org/abs/2309.08374
Spurred by the recent rapid increase in the development and distribution of large language models (LLMs) across industry and academia, much recent work has drawn attention to safety- and security-related threats and vulnerabilities of LLMs, including
Externí odkaz:
http://arxiv.org/abs/2308.12833
Autor:
Cheng, Yuk Shan, Dadi, Kamalesh, Mitchell, Toby, Thompson, Samantha, Piskunov, Nikolai, Wright, Lewis D., Gawith, Corin B. E., McCracken, Richard A., Reid, Derryck T.
Publikováno v:
NATURE COMMUNICATIONS 15:1466 (2024)
The characterization of Earth-like exoplanets and precision tests of cosmological models using next-generation telescopes such as the ELT will demand precise calibration of astrophysical spectrographs in the visible region, where stellar absorption l
Externí odkaz:
http://arxiv.org/abs/2306.13533
Autor:
Griffin, Lewis D, Kleinberg, Bennett, Mozes, Maximilian, Mai, Kimberly T, Vau, Maria, Caldwell, Matthew, Marvor-Parker, Augustine
Two studies tested the hypothesis that a Large Language Model (LLM) can be used to model psychological change following exposure to influential input. The first study tested a generic mode of influence - the Illusory Truth Effect (ITE) - where earlie
Externí odkaz:
http://arxiv.org/abs/2303.06074
Publikováno v:
PLoS ONE 18(8) (2023): e0285333
Speech deepfakes are artificial voices generated by machine learning models. Previous literature has highlighted deepfakes as one of the biggest security threats arising from progress in artificial intelligence due to their potential for misuse. Howe
Externí odkaz:
http://arxiv.org/abs/2301.07829
Adversarial examples in NLP are receiving increasing research attention. One line of investigation is the generation of word-level adversarial examples against fine-tuned Transformer models that preserve naturalness and grammaticality. Previous work
Externí odkaz:
http://arxiv.org/abs/2210.11598
Autor:
Mavor-Parker, Augustine N., Sargent, Matthew J., Pehle, Christian, Banino, Andrea, Griffin, Lewis D., Barry, Caswell
Reinforcement learning agents must painstakingly learn through trial and error what sets of state-action pairs are value equivalent -- requiring an often prohibitively large amount of environment experience. MDP homomorphisms have been proposed that
Externí odkaz:
http://arxiv.org/abs/2209.06356
Current deep learning methods for anomaly detection in text rely on supervisory signals in inliers that may be unobtainable or bespoke architectures that are difficult to tune. We study a simpler alternative: fine-tuning Transformers on the inlier da
Externí odkaz:
http://arxiv.org/abs/2204.05695