Lowest Rated Papers

Rating: Lowest Clear all

Advanced filtersTopic: All

Experimental

Topic

Topics are auto-detected from title, abstract, and metadata and may be imperfect.

Publication Date

Newest Oldest

Average Rating

Highest Lowest Clear

FoldSAE: Learning to Steer Protein Folding Through Sparse Representations

Wojciech Zarzecki, Paulina Szymczak, Ewa Szczurek, Kamil Deja

arXiv·2025

RFdiffusion is a popular and well-established model for generation of protein structures. However, this generative process offers limited insight into its internal representations and how they contribute to the final protein structure. Concurrently, recent work in mechanistic interpretability has successfully used Sparse Autoencoders (SAEs) to discover interpretable features within neural networks. We combine these concepts by applying SAE to the internal representations of RFdiffusion to uncover secondary structure-specific features and establish a relationship between them and generated protein structures. Building on these insights, we introduce a novel steering mechanism that enables precise control of secondary structure formation through a tunable hyperparameter, while simultaneously revealing interpretable block and neuron-level representations within RFdiffusion. Our work pioneers a new framework for making RFdiffusion more interpretable, demonstrating how understanding internal features can be directly translated into precise control over the protein design process.

No ratings yet

View paper →

Investigating Context-Driven Multimodal Tools to Support DHH Children's ASL Exposure at Home

James Spann

ACM SIGACCESS Accessibility and Computing·2025

Having caregivers actively engage in child-directed speech, and providing face-to-face communication to their child, can play a central role in the child's language, cognitive, and social development over their lifetime. However, over 90% of Deaf and Hard of Hearing (DHH) children are born to hearing parents who might not know American Sign Language (ASL) and are faced with the potentially steep learning curve of a new language. With my dissertation work, I plan on investigating assistive communication technologies that can support caregivers in communicating with their DHH children. This work aims to explore design guidelines and hands-on prototypes that can bridge this communicative gap.

No ratings yet

View paper →

Why Does Thinking Feel So Hard?

Carl Hendrick

Substack·2025

Effort Feels Costly When the Brain Can't Detect Progress

No ratings yet

View paper →

Understanding education policy preferences: Survey experiments with policymakers in 35 developing countries

Lee Crawfurd, Susannah Hares, Ana Minardi, Justin Sandefur

World Development

No abstract available

No ratings yet

View paper →

PreviousPage 148 of 467Next

Mysterious illnesses have supernatural and ritualistic cures: Evidence from 3,655 century-old Irish folk cures

Mícheál de Barra, Ángel V. Jiménez, Nachita Rosun, Aiyana K. Willard

Proceedings of the National Academy of Sciences·2025

Why and when do people draw upon religious and supernatural solutions to problems? Cognitive scientists and anthropologists have proposed a range of answers, stressing religion and ritual’s capacity to alleviate anxiety, create a sense of order, or explain otherwise inexplicable events. Here, we leverage a unique dataset of 3,655 folk cures for 35 diseases, collected in 1937/8 from a mostly rural Irish sample born roughly between 1850 and 1925. Since the diseases vary in theory-relevant ways and the cures vary in the degree to which they include religious and supernatural elements, this dataset facilitates a unique test of these predictions in a premodern western population. In preregistered tests, we find that diseases judged by two doctors to have causes and mechanisms that would be unclear to the patients were more likely to have supernatural/religious treatments. Contra common predictions, severe and disabling diseases did not have more supernatural/religious cures and anxiety-provoking diseases did not have more ritualistic cures.

No ratings yet

View paper →

Proof of a perfect platonic representation hypothesis

Liu Ziyin, Isaac Chuang

arXiv·2025

In this note, we elaborate on and explain in detail the proof given by Ziyin et al. (2025) of the ``perfect" Platonic Representation Hypothesis (PRH) for the embedded deep linear network model (EDLN). We show that if trained with the stochastic gradient descent (SGD), two EDLNs with different widths and depths and trained on different data will become Perfectly Platonic, meaning that every possible pair of layers will learn the same representation up to a rotation. Because most of the global minima of the loss function are not Platonic, that SGD only finds the perfectly Platonic solution is rather extraordinary. The proof also suggests at least six ways the PRH can be broken. We also show that in the EDLN model, the emergence of the Platonic representations is due to the same reason as the emergence of progressive sharpening. This implies that these two seemingly unrelated phenomena in deep learning can, surprisingly, have a common cause. Overall, the theory and proof highlight the importance of understanding emergent "entropic forces" due to the irreversibility of SGD training and their role in representation learning. The goal of this note is to be instructive while avoiding jargon and lengthy technical details.

No ratings yet

View paper →

Charming Witches:

Diane Purkiss

Preternature: Critical and Historical Studies on the Preternatural·2014

ABSTRACT This article examines the charms transcribed in detail in Thomas Potts's pamphlet The wonderfull discouerie of witches in the countie of Lancaster (1613), hitherto treated only briefly by historians. All the charms are related closely to authentic and well-attested older and mostly late medieval traditions of charming, identified here. After considering the possibility that Potts includes these charms as part of his anti-Papist polemical stance, the article then ponders the ways in which the charms are both meticulously detailed and also garbled. Through an examination of the role of gabbling or garbling in both medieval high magic and in Protestant anti-Latin satire, the place of garbled or ungrammatical Latin in particular is presented as an integral part both of magic and of antipopery. In particular, the reference to the Five Wounds in one of the charms is read as bringing together the themes of authentic remembrance of Catholic religious practice and fear of Catholic insurgency harking back to the rebellions of the past.

No ratings yet

View paper →

Getting It Wrong

Diane Purkiss

Pomegranate·2020·1 citations

This article is an examination of recent best-selling fictions and television adaptations which portray the history of witchcraft, often using outmoded historical theses, and often falsifying the known life histories of actual convicted witches. This article argues that these fictions, marked by problematically eugenicist ideas of magic, and in one case by a very uncomfortable appropriation of the Holocaust, are ultimately unhelpful to Pagans because they falsify history and deny the real needs of the contemporary Pagan communities.

No ratings yet

View paper →

Revisiting Small Batch Training for Deep Neural Networks

Dominic Masters, Carlo Luschi

arXiv·2018

Modern deep neural network training is typically based on mini-batch stochastic gradient optimization. While the use of large mini-batches increases the available computational parallelism, small batch training has been shown to provide improved generalization performance and allows a significantly smaller memory footprint, which might also be exploited to improve machine throughput. In this paper, we review common assumptions on learning rate scaling and training duration, as a basis for an experimental comparison of test performance for different mini-batch sizes. We adopt a learning rate that corresponds to a constant average weight update per gradient calculation (i.e., per unit cost of computation), and point out that this results in a variance of the weight updates that increases linearly with the mini-batch size $m$. The collected experimental results for the CIFAR-10, CIFAR-100 and ImageNet datasets show that increasing the mini-batch size progressively reduces the range of learning rates that provide stable convergence and acceptable test performance. On the other hand, small mini-batch sizes provide more up-to-date gradient calculations, which yields more stable and reliable training. The best performance has been consistently obtained for mini-batch sizes between $m = 2$ and $m = 32$, which contrasts with recent work advocating the use of mini-batch sizes in the thousands.

No ratings yet

View paper →

Train longer, generalize better: closing the generalization gap in large batch training of neural networks

Elad Hoffer, Itay Hubara, Daniel Soudry

arXiv·2017

Background: Deep learning models are typically trained using stochastic gradient descent or one of its variants. These methods update the weights using their gradient, estimated from a small fraction of the training data. It has been observed that when using large batch sizes there is a persistent degradation in generalization performance - known as the "generalization gap" phenomena. Identifying the origin of this gap and closing it had remained an open problem. Contributions: We examine the initial high learning rate training phase. We find that the weight distance from its initialization grows logarithmically with the number of weight updates. We therefore propose a "random walk on random landscape" statistical model which is known to exhibit similar "ultra-slow" diffusion behavior. Following this hypothesis we conducted experiments to show empirically that the "generalization gap" stems from the relatively small number of updates rather than the batch size, and can be completely eliminated by adapting the training regime used. We further investigate different techniques to train models in the large-batch regime and present a novel algorithm named "Ghost Batch Normalization" which enables significant decrease in the generalization gap without increasing the number of updates. To validate our findings we conduct several additional experiments on MNIST, CIFAR-10, CIFAR-100 and ImageNet. Finally, we reassess common practices and beliefs concerning training of deep models and suggest they may not be optimal to achieve good generalization.

No ratings yet

View paper →

Lowest Rated Papers

FoldSAE: Learning to Steer Protein Folding Through Sparse Representations

Investigating Context-Driven Multimodal Tools to Support DHH Children's ASL Exposure at Home

Why Does Thinking Feel So Hard?

Understanding education policy preferences: Survey experiments with policymakers in 35 developing countries

Mysterious illnesses have supernatural and ritualistic cures: Evidence from 3,655 century-old Irish folk cures

Proof of a perfect platonic representation hypothesis

Arcee AI | Arcee Debuts Trinity Mini, Expanding Its U.S.-Built Model Line

Charming Witches:

Project MUSE - English Witches and SS Academics: Evaluating Sources for the English Witch Trials in Himmler's Hexenkartothek

Getting It Wrong

Revisiting Small Batch Training for Deep Neural Networks

Train longer, generalize better: closing the generalization gap in large batch training of neural networks