Zobrazeno 1 - 10
of 5 700
pro vyhledávání: '"Gallé, A."'
Autor:
Khalifa, Muhammad, Tan, Yi-Chern, Ahmadian, Arash, Hosking, Tom, Lee, Honglak, Wang, Lu, Üstün, Ahmet, Sherborne, Tom, Gallé, Matthias
Model merging has shown great promise at combining expert models, but the benefit of merging is unclear when merging ``generalist'' models trained on many tasks. We explore merging in the context of large ($\sim100$B) models, by \textit{recycling} ch
Externí odkaz:
http://arxiv.org/abs/2412.04144
Autor:
Zhao, Wenting, Jiang, Nan, Lee, Celine, Chiu, Justin T, Cardie, Claire, Gallé, Matthias, Rush, Alexander M
With the goal of benchmarking generative systems beyond expert software development ability, we introduce Commit0, a benchmark that challenges AI agents to write libraries from scratch. Agents are provided with a specification document outlining the
Externí odkaz:
http://arxiv.org/abs/2412.01769
Autor:
Daniault, Louis, Kaur, Jaismeen, Gallé, Geoffrey, Sire, Cedric, Sylla, FrançOis, Lopez-Martens, Rodrigo
We report on the nonlinear temporal post-compression of 7 mJ sub-40 fs pulses from a commercial kHz Ti:Sapphire laser down to a record 4 fs duration (1.5 optical cycle) in a compact single-stage gas-filled multipass cell, with 60% overall compression
Externí odkaz:
http://arxiv.org/abs/2410.12525
Autor:
Matton, Alexandre, Sherborne, Tom, Aumiller, Dennis, Tommasone, Elena, Alizadeh, Milad, He, Jingyi, Ma, Raymond, Voisin, Maxime, Gilsenan-McMahon, Ellen, Gallé, Matthias
In this paper, we consider contamination by code generation test sets, in particular in their use in modern large language models. We discuss three possible sources of such contamination and show findings supporting each of them: (i) direct data leak
Externí odkaz:
http://arxiv.org/abs/2407.07565
Autor:
Ye, Zihuiwen, Greenlee-Scott, Fraser, Bartolo, Max, Blunsom, Phil, Campos, Jon Ander, Gallé, Matthias
Reward models (RMs) play a critical role in aligning language models through the process of reinforcement learning from human feedback. RMs are trained to predict a score reflecting human preference, which requires significant time and cost for human
Externí odkaz:
http://arxiv.org/abs/2405.20850
Autor:
Casetti-Dinescu, Dana I., Baena-Galle, Roberto, Girard, Terrence M., Cervantes-Rovira, Alejandro, Todeasa, Sebastian
We present an expanded and improved deep-learning (DL) methodology for determining centers of star images on HST/WFPC2 exposures. Previously, we demonstrated that our DL model can eliminate the pixel-phase bias otherwise present in these undersampled
Externí odkaz:
http://arxiv.org/abs/2404.16995
Humans follow criteria when they execute tasks, and these criteria are directly used to assess the quality of task completion. Therefore, having models learn to use criteria to provide feedback can help humans or models to perform tasks better. Howev
Externí odkaz:
http://arxiv.org/abs/2403.01069
Autor:
Ahmadian, Arash, Cremer, Chris, Gallé, Matthias, Fadaee, Marzieh, Kreutzer, Julia, Pietquin, Olivier, Üstün, Ahmet, Hooker, Sara
AI alignment in the shape of Reinforcement Learning from Human Feedback (RLHF) is increasingly treated as a crucial ingredient for high performance large language models. Proximal Policy Optimization (PPO) has been positioned by recent literature as
Externí odkaz:
http://arxiv.org/abs/2402.14740
Extractive summaries are usually presented as lists of sentences with no expected cohesion between them. In this paper, we aim to enforce cohesion whilst controlling for informativeness and redundancy in summaries, in cases where the input exhibits h
Externí odkaz:
http://arxiv.org/abs/2402.10643
Publikováno v:
Basic and Applied Ecology, Vol 80, Iss , Pp 92-100 (2024)
Climate change negatively affects arthropod biodiversity worldwide. Mitigating the resulting arthropod decline is a great challenge. Dwarf shrubs in open areas might buffer microclimatic extremities by reducing the solar radiation reaching the ground
Externí odkaz:
https://doaj.org/article/56fd1cc00ead447e9dc89926a56a0ae6
