Teaching LLMs to reason like Bayesians
Generative AI
Like
0
Liked
Liked
digitado
About digitado
Posts by :
Lawsuit: Google Gemini sent man on violent missions, set suicide “countdown”
A man killed himself after the Google Gemini chatbot pushed him to kill innocent strangers and then started a countdown for the man to take his own life, a wrongful-death lawsuit filed against Google by the man’s father alleged. “In the days leading up to his death, Jonathan Gavalas was trapped in a collapsing reality built by Google’s Gemini chatbot,” said the lawsuit filed today in US District Court for the Northern District of California. “Gemini convinced him […]
Like
0
Liked
Liked
Why Do Neural Networks Forget: A Study of Collapse in Continual Learning
Catastrophic forgetting is a major problem in continual learning, and lots of approaches arise to reduce it. However, most of them are evaluated through task accuracy, which ignores the internal model structure. Recent research suggests that structural collapse leads to loss of plasticity, as evidenced by changes in effective rank (eRank). This indicates a link to forgetting, since the networks lose the ability to expand their feature space to learn new tasks, which forces the network to overwrite […]
Like
0
Liked
Liked
Fusion and Grouping Strategies in Deep Learning for Local Climate Zone Classification of Multimodal Remote Sensing Data
Local Climate Zones (LCZs) give a zoning map to study urban structures and land use and analyze the impact of urbanization on local climate. Multimodal remote sensing enables LCZ classification, for which data fusion is significant for improving accuracy owing to the data complexity. However, there is a gap in a comprehensive analysis of the fusion mechanisms used in their deep learning (DL) classifier architectures. This study analyzes different fusion strategies in the multi-class LCZ classification models for […]
Like
0
Liked
Liked
Weather-Related Crash Risk Forecasting: A Deep Learning Approach for Heterogenous Spatiotemporal Data
This study introduces a deep learning-based framework for forecasting weather-related traffic crash risk using heterogeneous spatiotemporal data. Given the complex, non-linear relationship between crash occurrence and factors such as road characteristics, and traffic conditions, we propose an ensemble of Convolutional Long Short-Term Memory (ConvLSTM) models trained over overlapping spatial grids. This approach captures both spatial dependencies and temporal dynamics while addressing spatial heterogeneity in crash patterns. North Carolina was selected as the study area due to its diverse […]
Like
0
Liked
Liked
Oracle-efficient Hybrid Learning with Constrained Adversaries
The Hybrid Online Learning Problem, where features are drawn i.i.d. from an unknown distribution but labels are generated adversarially, is a well-motivated setting positioned between statistical and fully-adversarial online learning. Prior work has presented a dichotomy: algorithms that are statistically-optimal, but computationally intractable (Wu et al., 2023), and algorithms that are computationally-efficient (given an ERM oracle), but statistically-suboptimal (Wu et al., 2024). This paper takes a significant step towards achieving statistical optimality and computational efficiency simultaneously in the […]
Like
0
Liked
Liked
After a rocky six years, Sony cancels future single-player PC game releases
Sony no longer plans to bring current and future single-player games to personal computers, according to Bloomberg. The report specifically names last year’s Ghost of Yotei and the soon-to-be-released Returnal successor, Saros, as games whose PC plans have been canceled. Some multiplayer and third-party titles will still reach PCs, however. Bloomberg’s Jason Schreier cites “people familiar with the company’s plans,” who say that some within the company worry that releasing the games on PC could hurt sales of […]
Like
0
Liked
Liked
Projected Hessian Learning: Fast Curvature Supervision for Accurate Machine-Learning Interatomic Potentials
The Hessian matrix (second derivatives) encodes far richer local curvature of the potential energy surface than energies and forces alone. However, training machine-learning interatomic potentials (MLIPs) with full Hessians is often impractical because explicitly forming and storing Hessian matrices scales quadratically in cost and memory. We introduce Projected Hessian Learning (PHL), a scalable second-order training framework that injects curvature information using only Hessian-vector products (HVPs). Rather than constructing the Hessian, PHL projects curvature along stochastic probe directions and […]
Like
0
Liked
Liked
Re-creating the complex cuisine of prehistoric Europeans
Archaeologists are keen to learn more about the specific diets and culinary practices of ancient populations around the globe. An interdisciplinary team of scientists analyzed the residues on prehistoric ceramic cooking pots and concluded that early Eastern European hunter-gatherer-fishers likely foraged for plants as well as hunted fish and other animals for their sustenance, according to a new paper published in the journal PLoS ONE. And they often combined ingredients for region-specific recipes. This is a burgeoning area […]
Like
0
Liked
Liked
SELDON: Supernova Explosions Learned by Deep ODE Networks
The discovery rate of optical transients will explode to 10 million public alerts per night once the Vera C. Rubin Observatory’s Legacy Survey of Space and Time comes online, overwhelming the traditional physics-based inference pipelines. A continuous-time forecasting AI model is of interest because it can deliver millisecond-scale inference for thousands of objects per day, whereas legacy MCMC codes need hours per object. In this paper, we propose SELDON, a new continuous-time variational autoencoder for panels of sparse […]
Like
0
Liked
Liked