Tootfinder

Rams' Puka Nacua feels like 'kid in the candy store' getting to learn from Davante Adams https://www.nfl.com/news/rams-puka-nacua-feels-like-kid-in-the-candy-store-getting-to-learn-from-davante-adams

Low-rank Momentum Factorization for Memory Efficient Training
Pouria Mahdavinia, Mehrdad Mahdavi
https://arxiv.org/abs/2507.08091 https://arxiv.org/pdf/2507.08091 https://arxiv.org/html/2507.08091
arXiv:2507.08091v1 Announce Type: new
Abstract: Fine-tuning large foundation models presents significant memory challenges due to stateful optimizers like AdamW, often requiring several times more GPU memory than inference. While memory-efficient methods like parameter-efficient fine-tuning (e.g., LoRA) and optimizer state compression exist, recent approaches like GaLore bridge these by using low-rank gradient projections and subspace moment accumulation. However, such methods may struggle with fixed subspaces or computationally costly offline resampling (e.g., requiring full-matrix SVDs). We propose Momentum Factorized SGD (MoFaSGD), which maintains a dynamically updated low-rank SVD representation of the first-order momentum, closely approximating its full-rank counterpart throughout training. This factorization enables a memory-efficient fine-tuning method that adaptively updates the optimization subspace at each iteration. Crucially, MoFaSGD leverages the computed low-rank momentum factors to perform efficient spectrally normalized updates, offering an alternative to subspace moment accumulation. We establish theoretical convergence guarantees for MoFaSGD, proving it achieves an optimal rate for non-convex stochastic optimization under standard assumptions. Empirically, we demonstrate MoFaSGD's effectiveness on large language model alignment benchmarks, achieving a competitive trade-off between memory reduction (comparable to LoRA) and performance compared to state-of-the-art low-rank optimization methods. Our implementation is available at https://github.com/pmahdavi/MoFaSGD.
toXiv_bot_toot

Sensitive infrared surface photovoltage in quasi-equilibrium in a layered semiconductor at low-intensity low-temperature condition
Qiang Wan, Keming Zhao, Guohao Dong, Enting Li, Tianyu Yang, Hao Wang, Yaobo Huang, Yao Wen, Yiwei Li, Jun He, Youguo Shi, Hong Ding, Nan Xu
https://arxiv.org/abs/2507.08279

Sensitive infrared surface photovoltage in quasi-equilibrium in a layered semiconductor at low-intensity low-temperature condition
Benefit to layer-dependent bandgap, van der Waals materials with surface photovoltaic effect (SPV) enable photodetection over a tunable wavelength range with low power consumption. However, sensitive SPV in the infrared region, especially in a quasi-steady illumination condition, is still elusive in layered semiconductors. Here, using angle-resolved photoemission spectroscopy, we report a sensitive SPV in quasi-equilibrium in NbSi0.5Te2, with photoresponsivity up to 2.4*10^6 V/(W*cm^(-2)) at lo…

Highly accurate simulations of asymmetric black-hole scattering and cross validation of effective-one-body models
Oliver Long, Harald P. Pfeiffer, Alessandra Buonanno, Gustav Uhre Jakobsen, Gustav Mogull, Antoni Ramos-Buades, Hannes R. R\"uter, Lawrence E. Kidder, Mark A. Scheel
https://arxiv.org/abs/2507.08071

Highly accurate simulations of asymmetric black-hole scattering and cross validation of effective-one-body models
The study of unbound binary-black-hole encounters provides a gauge-invariant approach to exploring strong-field gravitational interactions in two-body systems, which can subsequently inform waveform models for bound orbits. In this work, we present 60 new highly accurate numerical relativity (NR) simulations of black-hole scattering, generated using the Spectral Einstein Code (SpEC). Our simulations include 14 spin-aligned configurations, as well as 16 configurations with unequal masses, up to …

Acting FAA Administrator Chris Rocheleau told the House Appropriations Committee that the Federal Aviation Administration plans to replace its aging air traffic control systems
-- which still rely on floppy disks and Windows 95 computers!
The agency has issued a Request For Information to gather proposals from companies willing to tackle the massive infrastructure overhaul

US air traffic control still runs on Windows 95 and floppy disks
Agency seeks contractors to modernize decades-old systems within four years.

Massively parallel and universal approximation of nonlinear functions using diffractive processors
Md Sadman Sakib Rahman, Yuhang Li, Xilin Yang, Shiqi Chen, Aydogan Ozcan
https://arxiv.org/abs/2507.08253

Massively parallel and universal approximation of nonlinear functions using diffractive processors
Nonlinear computation is essential for a wide range of information processing tasks, yet implementing nonlinear functions using optical systems remains a challenge due to the weak and power-intensive nature of optical nonlinearities. Overcoming this limitation without relying on nonlinear optical materials could unlock unprecedented opportunities for ultrafast and parallel optical computing systems. Here, we demonstrate that large-scale nonlinear computation can be performed using linear optics…

Filing: neobank Chime plans to sell 26M shares in its IPO at $24 to $26, giving it a valuation between $10.3B and $11.1B; its two co-founders own 4% to 5% each (Cory Weinberg/The Information)
https://www.theinformation.com/briefings/chime-sets-ipo-…

Chime Sets Tentative IPO Price at Around $10.7 Billion
Banking app Chime plans to sell shares in an initial public offering at a price that would value it at between $10.3 billion and $11.1 billion, on a fully-diluted basis, the company said in a new securities filing Monday. The company set the price range ahead of a roadshow to meet investors over the next week, ahead of its IPO. The per-share price of the range, between $24 and $26, is more

🇺🇦 #NowPlaying on #BBC6Music's #LaurenLaverne
Unknown Mortal Orchestra:
🎵 Swim and Sleep (Like A Shark)
#UnknownMortalOrchestra
https://unknown-mortal-orchestra.bandcamp.com/track/swim-and-sleep-like-a-shark-lindstr-m-remix
https://open.spotify.com/track/265ehI4I7NfR8PmAXdpspn

In last week's roadmap update, I mentioned that we were just about ready for folks to take OmniFocus 4.7 through its paces in public test builds.
Well, now we're ready! The OmniFocus 4.7 public test introduces Planned dates, mutually exclusive tags, repeat counts and end dates, time-sensitive notifications, and more.
We look forward to your feedback!

A model-agnostic gravitational-wave background characterization algorithm
Taylor Knapp, Patrick M. Meyers, Arianna I. Renzini
https://arxiv.org/abs/2507.08095

A model-agnostic gravitational-wave background characterization algorithm
As ground-based gravitational-wave (GW) detectors improve in sensitivity, gravitational-wave background (GWB) signals will progressively become detectable. Currently, searches for the GWB model the signal as a power law, however deviations from this model will be relevant at increased sensitivity. Therefore, to prepare for the range of potentially detectable GWB signals, we propose an interpolation model implemented through a transdimensional reverse-jump Markov chain Monte Carlo (RJMCMC) algor…