Search the full wire by company, model, lab, or keyword. Every story we have ever aggregated.
Tabular foundation models perform molecular property prediction via in-context learning without task-specific fine-tuning, matching or exceeding classical baselines on small datasets.
Univariate Channel Fusion (UCF) reduces computational cost for multivariate time series classification, enabling deployment on IoT and wearable devices.
SCRIPT is an adaptive intelligent tutoring system for Python programming leveraging generative models, designed for university-level teaching and research.
Researchers distilled reasoning systems from humanities scholars' published work into LLM inference constraints to automate doctoral supervision, peer review, and lecturing tasks.
Theoretical analysis establishes sample complexity lower bounds for stochastic shortest path problems with generative models, revealing challenges when minimum cost approaches zero.
Critical examination of AI accountability frameworks argues industry stakeholders use rhetorical decoys to obscure power consolidation and political economy dynamics.
Proposes multi-modal CNN framework combining CT radiology and histopathology imaging for improved lung cancer diagnosis and interpretability.
Anthropic Labs launches Claude Design, enabling collaborative visual content creation including designs, prototypes, and slides.
Addresses correlated device failure in federated learning by extending Probabilistic Synchronous Parallel sampling to handle dependent node availability patterns.
Analyzes convergence rates for uncoupled learning algorithms in zero-sum games with bandit feedback, proving improved bounds vs. prior work.
Proposes self-supervised learning variant that preserves appearance-based features for tasks like weather analysis where visual style is discriminative.
Introduces probabilistic traffic forecasting by replacing output layers with Gaussian Mixture Models to capture uncertainty in spatio-temporal predictions.
Applies YOLOv12 object detection to multi-class acute myeloid leukemia cell classification using segmentation preprocessing techniques.
Introduces Prototype-Grounded Concept Models that ground learned concepts in visual prototypes to enable verification and human correction of concept alignment.
Theoretical work on linear system solver convergence via backward error analysis; no direct AI applications.
Google travel planning tools using Gemini; consumer product announcement without technical detail.
AEGIS method preserves vision-language knowledge during robotic control fine-tuning via gradient isolation.
Approximation algorithms for fair clustering with demographic constraints; theoretical CS work without LLM focus.
Chain-of-thought prompting degrades spatial reasoning in multimodal LLMs; identifies shortcut learning and hallucination.
LLMSniffer detects LLM-generated code using GraphCodeBERT and contrastive learning on GPTSniffer/Whodunit datasets.
AST: training-free speech editing framework using latent recomposition from pretrained TTS models.
Mind's Eye benchmark evaluates multimodal LLMs on visual abstraction, relation, transformation via A-R-T taxonomy.
Theoretical framework for Hebbian plasticity using Wasserstein geometry; neuroscience formalism without ML applications.
In the latest sign of AI silly season, Allbirds, the shoe company, told the world it was now an AI company and briefly managed to septuple its stock price. The Newbird AI story is really just one of a bunch of things this week that made us wonder: have we reached the peak of AI, or at least a peak of AI? Verge subscribers, don't forget you get exclusive access to ad-free Vergecast wherever you get your podcasts. Head here. Not a subscriber? You can sign up here. On this episode of The Vergecast, we look at both the data and the vibes. David and Nilay explore a new study from Stanford that say...
Embedded system for measuring ambulance vibrations to minimize patient cabin movement; not AI-relevant.
Survey of intrinsic interpretability architectures for LLMs; systematizes design principles for transparent model internals.
The gap between AI insiders and everyone else is widening, and the spending, suspicion, and even new vocabulary are starting to show it. While OpenAI is busy buying up everything from finance apps to talk shows, a certain shoe company just rebranded as an AI infrastructure play, and Anthropic unveiled a model it says is too powerful to release publicly …but apparently not too […]
Roboticists used to dream big but build small. They’d hope to match or exceed the extraordinary complexity of the human body, and then they’d spend their career refining robotic arms for auto plants. Aim for C-3P0; end up with the Roomba. The real ambition for many of these researchers was the robot of science fiction—one…
Datasette 1.0a28 alpha release fixes compatibility bugs in execute_write_fn callbacks and adds database close methods.
Anthropic releases Claude Opus 4.7, claiming state-of-the-art performance across all dimensions versus version 4.6.