Zobrazeno 1 - 10
of 10
pro vyhledávání: '"Scherlis, Adam"'
We propose affine concept editing (ACE) as an approach for steering language models' behavior by intervening directly in activations. We begin with an affine decomposition of model activation vectors and show that prior methods for steering model beh
Externí odkaz:
http://arxiv.org/abs/2411.09003
Individual neurons in neural networks often represent a mixture of unrelated features. This phenomenon, called polysemanticity, can make interpreting neural networks more difficult and so we aim to understand its causes. We propose doing so through t
Externí odkaz:
http://arxiv.org/abs/2210.01892
Autor:
Ziegler, Daniel M., Nix, Seraphina, Chan, Lawrence, Bauman, Tim, Schmidt-Nielsen, Peter, Lin, Tao, Scherlis, Adam, Nabeshima, Noa, Weinstein-Raun, Ben, de Haas, Daniel, Shlegeris, Buck, Thomas, Nate
In the future, powerful AI systems may be deployed in high-stakes settings, where a single failure could be catastrophic. One technique for improving AI safety in high-stakes settings is adversarial training, which uses an adversary to generate examp
Externí odkaz:
http://arxiv.org/abs/2205.01663
Recently, the problem of unitarity violation during the preheating stage of Higgs inflation with a large non-minimal coupling has been much discussed in the literature. We point out that this problem can be translated into a strong coupling problem f
Externí odkaz:
http://arxiv.org/abs/2007.04701
Autor:
Fort, Stanislav, Scherlis, Adam
We explore the loss landscape of fully-connected and convolutional neural networks using random, low-dimensional hyperplanes and hyperspheres. Evaluating the Hessian, $H$, of the loss function on these hypersurfaces, we observe 1) an unusual excess o
Externí odkaz:
http://arxiv.org/abs/1807.02581
Autor:
Graham, Peter W., Scherlis, Adam
Publikováno v:
Phys. Rev. D 98, 035017 (2018)
For the minimal QCD axion model it is generally believed that overproduction of dark matter constrains the axion mass to be above a certain threshold, or at least that the initial misalignment angle must be tuned if the mass is below that threshold.
Externí odkaz:
http://arxiv.org/abs/1805.07362
Autor:
Scherlis, Adam
The classification of Grassmannian cluster algebras resembles that of regular polygonal tilings. We conjecture that this resemblance may indicate a deeper connection between these seemingly unrelated structures.
Comment: 3 pages, 2 tables, 2 fig
Comment: 3 pages, 2 tables, 2 fig
Externí odkaz:
http://arxiv.org/abs/1510.07777
Multi-loop scattering amplitudes in N=4 Yang-Mills theory possess cluster algebra structure. In order to develop a computational framework which exploits this connection, we show how to construct bases of Goncharov polylogarithm functions, at any wei
Externí odkaz:
http://arxiv.org/abs/1507.01950
Akademický článek
Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.
K zobrazení výsledku je třeba se přihlásit.
Akademický článek
Tento výsledek nelze pro nepřihlášené uživatele zobrazit.
K zobrazení výsledku je třeba se přihlásit.
K zobrazení výsledku je třeba se přihlásit.