Tootfinder

Libra: Assessing and Improving Reward Model by Learning to Think
Meng Zhou, Bei Li, Jiahao Liu, Xiaowen Shi, Yang Bai, Rongxiang Weng, Jingang Wang, Xunliang Cai
https://arxiv.org/abs/2507.21645

Libra: Assessing and Improving Reward Model by Learning to Think
Reinforcement learning (RL) has significantly improved the reasoning ability of large language models. However, current reward models underperform in challenging reasoning scenarios and predominant RL training paradigms rely on rule-based or reference-based rewards, which impose two critical limitations: 1) the dependence on finely annotated reference answer to attain rewards; and 2) the requirement for constrained output format. These limitations fundamentally hinder further RL data scaling an…

HairCUP: Hair Compositional Universal Prior for 3D Gaussian Avatars
Byungjun Kim, Shunsuke Saito, Giljoo Nam, Tomas Simon, Jason Saragih, Hanbyul Joo, Junxuan Li
https://arxiv.org/abs/2507.19481

HairCUP: Hair Compositional Universal Prior for 3D Gaussian Avatars
We present a universal prior model for 3D head avatars with explicit hair compositionality. Existing approaches to build generalizable priors for 3D head avatars often adopt a holistic modeling approach, treating the face and hair as an inseparable entity. This overlooks the inherent compositionality of the human head, making it difficult for the model to naturally disentangle face and hair representations, especially when the dataset is limited. Furthermore, such holistic models struggle to su…

Nvidia-Chef Jensen Huang trifft Bundeskanzler
Im Rahmen einer ausgedehnten Europatour plant Huang auch ein Gespräch mit Friedrich Merz. Es dürfte um europäische KI-Rechenzentren gehen.
https://www.

Further aspects of Supersymmetric Virasoro Minimal Strings
Clifford V. Johnson
https://arxiv.org/abs/2506.19000 https://arxiv.org/pdf…

Further aspects of Supersymmetric Virasoro Minimal Strings
A supersymmetric version of the Virasoro minimal string was defined some time ago using random matrix model techniques. Several of the special properties of the matrix model that were noted (accessible fully non-perturbatively) underlie features shared by both 0A and 0B versions of the theory. This paper develops the formalism much further, including showing how the 0A and 0B choices familiar in continuum approaches have a direct analogue in terms of building solutions of the appropriate string…

🔍 Vortrag: Die Mathematik des jungen Carl Friedrich #Gauß
Der Vortrag gibt einige Einblicke in die mathematischen Interessen und Resultate des jungen Gauß bis zu seiner Dissertation.
⏲ Uhrzeit: 19:00 & 20:00 Uhr
🗓️ Datum: 21.06.2025
⏱ Dauer: 30min
🧑‍🔬 Veranstalter: Institut für Numerische und Angewandte Mathematik
📍 Veranstaltungsort: ZHG004, ZHG, Platz…

Pragmatic Policy Development via Interpretable Behavior Cloning
Anton Matsson, Yaochen Rao, Heather J. Litman, Fredrik D. Johansson
https://arxiv.org/abs/2507.17056

Pragmatic Policy Development via Interpretable Behavior Cloning
Offline reinforcement learning (RL) holds great promise for deriving optimal policies from observational data, but challenges related to interpretability and evaluation limit its practical use in safety-critical domains. Interpretability is hindered by the black-box nature of unconstrained RL policies, while evaluation -- typically performed off-policy -- is sensitive to large deviations from the data-collecting behavior policy, especially when using methods based on importance sampling. To add…

On geometry of $Q^{(2k)}_g$-curvature
Mingxiang Li, Juncheng Wei, Xingwang Xu
https://arxiv.org/abs/2506.20165 https://arxiv.org/pdf/…

On geometry of $Q^{(2k)}_g$-curvature
The main purpose of current article is to study the geometry of $Q$-curvature. For simplicity, we start with a very simple model: a complete and conformal metric $g=e^{2u}|dx|^2$ on $\mathbb{R}^n$. We assume the metric $g$ has non-negative $nth$-order $Q$-curvature and non-negative scalar curvature. We show that the Ricci curvature is non-negative. We further assume that the metric $g$ is not degenerate in the sense that the isoperimetric ratio near the end is positive. With this extr…

"There is no reason to believe [rent controls] have any effect on new housing construction. One could go a step further: economically, land use reforms and stronger rent regulations should go together."
https://jwmason.org/slackwire/can-zohran-do-it/

Ultrastable, low-error dynamic polarization encoding of deterministically generated single photons
Joscha Hanel, Zenghui Jiang, Jipeng Wang, Frederik Benthin, Tom Fandrich, Eddy Patrick Rugeramigabo, Raphael Joos, Michael Jetter, Simone Luca Portalupi, Jingzhong Yang, Michael Zopf, Peter Michler, Fei Ding
https://arxiv.org/abs/2…

Ultrastable, low-error dynamic polarization encoding of deterministically generated single photons
The ability to inscribe information on single photons at high speeds is a crucial requirement for quantum applications such as quantum communication and measurement-based photonic quantum computation. Nowadays, most experimental implementations employ phase modulators in single-pass, Mach-Zehnder interferometer or Michelson interferometer configurations to encode information on photonic qubits. However, these approaches are intrinsically sensitive to environmental influences, limiting the achie…

Künstliche Intelligenz "Made in Germany". Bei einem Treffen zwischen Nvidia-CEO Jensen Huang und Bundeskanzler Friedrich Merz wurde eine wegweisende Zusammenarbeit verkündet.
Zum Artikel: https://heise.de/-10446867?wt_mc=sm.re