Tiny, long-armed dinosaur leads to rethink of dinosaur miniaturization

Tiny, long-armed dinosaur leads to rethink of dinosaur miniaturization

⋅

Alvarezsaurids were mostly small-bodied theropods that paleontologists originally misinterpreted as early flightless birds, only to later recognize them as an ant-eating lineage of non-avian dinosaurs. For years, we suspected that Alvarezsaurids underwent a rare process of evolutionary miniaturization directly coupled to a diet of social insects like ants and termites. It was a tidy hypothesis: They got smaller to become more efficient at catching ants. Now, a recently discovered fossil of one of the smallest alvarezsaurids ever found […]

Ver mais

Liked Liked
Hippotorch: Hippocampus-inspired episodic memory for sparse-reward problems

Hippotorch: Hippocampus-inspired episodic memory for sparse-reward problems

⋅

![img](socqna2mb7eg1) I’ve been working on a replay buffer replacement inspired by how the hippocampus consolidates memories during sleep. The problem: In sparse-reward tasks with long horizons (e.g., T-maze variants), the critical observation arrives at t=0 but the decision happens 30+ steps later. Uniform replay treats all transitions equally, so the rare successes get drowned out. The approach: Hippotorch uses a dual encoder to embed experiences, stores them in an episodic memory with semantic indices, and periodically runs a […]

Ver mais

Liked Liked
The rise of Moltbook suggests viral AI prompts may be the next big security threat

The rise of Moltbook suggests viral AI prompts may be the next big security threat

⋅

On November 2, 1988, graduate student Robert Morris released a self-replicating program into the early Internet. Within 24 hours, the Morris worm had infected roughly 10 percent of all connected computers, crashing systems at Harvard, Stanford, NASA, and Lawrence Livermore National Laboratory. The worm exploited security flaws in Unix systems that administrators knew existed but had not bothered to patch. Morris did not intend to cause damage. He wanted to measure the size of the Internet. But a […]

Ver mais

Liked Liked
MIT-IBM Watson AI Lab seed to signal: Amplifying early-career faculty impact

MIT-IBM Watson AI Lab seed to signal: Amplifying early-career faculty impact

⋅

The early years of faculty members’ careers are a formative and exciting time in which to establish a firm footing that helps determine the trajectory of researchers’ studies. This includes building a research team, which demands innovative ideas and direction, creative collaborators, and reliable resources.  For a group of MIT faculty working with and on artificial intelligence, early engagement with the MIT-IBM Watson AI Lab through projects has played an important role helping to promote ambitious lines of […]

Ver mais

Liked Liked

Learning to Remember, Learn, and Forget in Attention-Based Models

⋅

In-Context Learning (ICL) in transformers acts as an online associative memory and is believed to underpin their high performance on complex sequence processing tasks. However, in gated linear attention models, this memory has a fixed capacity and is prone to interference, especially for long sequences. We propose Palimpsa, a self-attention model that views ICL as a continual learning problem that must address a stability-plasticity dilemma. Palimpsa uses Bayesian metaplasticity, where the plasticity of each attention state is tied […]

Ver mais

Liked Liked

What Matters for Simulation to Online Reinforcement Learning on Real Robots

⋅

arXiv:2602.20220v1 Announce Type: new Abstract: We investigate what specific design choices enable successful online reinforcement learning (RL) on physical robots. Across 100 real-world training runs on three distinct robotic platforms, we systematically ablate algorithmic, systems, and experimental decisions that are typically left implicit in prior work. We find that some widely used defaults can be harmful, while a set of robust, readily adopted design choices within standard RL practice yield stable learning across tasks and hardware. These results […]

Ver mais

Liked Liked

Automated HER2 scoring with uncertainty quantification using lensfree holography and deep learning

⋅

Accurate assessment of human epidermal growth factor receptor 2 (HER2) expression is critical for breast cancer diagnosis, prognosis, and therapy selection; yet, most existing digital HER2 scoring methods rely on bulky and expensive optical systems. Here, we present a compact and cost-effective lensfree holography platform integrated with deep learning for automated HER2 scoring of immunohistochemically stained breast tissue sections. The system captures lensfree diffraction patterns of stained HER2 tissue sections under RGB laser illumination and acquires complex field […]

Ver mais

Liked Liked

Why Every Analytics Engineer Needs to Understand Data Architecture

⋅

Get the data architecture right, and everything else becomes easier. I know it sounds simple, but in reality, little nuances in designing your data architecture may have costly implications. This article provides a crash course on the architectures that shape your daily decisions – from relational databases to event-driven systems. The post Why Every Analytics Engineer Needs to Understand Data Architecture appeared first on Towards Data Science.

Ver mais

Liked Liked

Dirichlet-Neumann Waveform Relaxation Method with Multiple Subdomains for Reaction-Diffusion Equation with a Time Delay

⋅

arXiv:2602.12409v1 Announce Type: new Abstract: In this study, we present the numerical investigation of the Dirichlet-Neumann Waveform Relaxation (DNWR) algorithm applied to multiple subdomains for the reaction-diffusion equation with time delay. Various arrangements of transmission conditions between subdomains are explored and a series of numerical experiments are conducted to evaluate and compare the efficiency and effectiveness of these configurations.

Ver mais

Liked Liked

Grokking as a Phase Transition between Competing Basins: a Singular Learning Theory Approach

⋅

Grokking, the abrupt transition from memorization to generalisation after extended training, suggests the presence of competing solution basins with distinct statistical properties. We study this phenomenon through the lens of Singular Learning Theory (SLT), a Bayesian framework that characterizes the geometry of the loss landscape via the local learning coefficient (LLC), a measure of the local degeneracy of the loss surface. SLT links lower-LLC basins to higher posterior mass concentration and lower expected generalisation error. Leveraging this theory, […]

Ver mais

Liked Liked

Load More

↑

Digitado © 2025