Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers from the Technical University of Munich introduced GaussianGPT, a transformer-based model that generates 3D scenes autoregressively via next-token prediction on Gaussian splatting tokens, accepted as an oral …
X-Square Robot released WALL-SS, an autoregressive world model designed to close the simulation-to-reality gap in robot training, achieving an action sensitivity score of 0.290 compared to Cosmos3's 0.044. The model uses…
A developer ran a 48-hour stability probe on a free AI model, finding that output drift and stale-label echoing were more telling than accuracy benchmarks. The experiment, conducted on MonkeyCode's free server, revealed …
NVIDIA CUDA and TensorRT can significantly accelerate Physical AI workloads on NVIDIA hardware, according to a developer from v-modal. The developer explains that optimizing the complete data path, rather than focusing o…
Hugging Face's Pollen Robotics team opened pre-orders on August 27, 2026, for Microduck, a $399 open-source 25 cm bipedal robot whose movements are neural policies trained in a physics simulator and exported to hardware.…
Firecrawl has released pdf-inspector, an open-source library that classifies PDFs as text-based, scanned, image-based, or mixed, and routes only the pages that need OCR to the OCR engine. The tool, available in Python, N…
Xpeng unveiled its new VLA 6.3.0 model, which can predict driving scenarios up to 6 seconds ahead, during a Physical AI sharing session. The model features a 30-second memory window, a 300% boost in decision-making speed…
Researchers from UC Berkeley and MIT released FreeToken, an open-source inference engine that enables frontier Mixture-of-Experts (MoE) models to run on consumer hardware via dynamic co-execution, achieving 3–4x faster d…
Researchers have developed object-agnostic grasp planners that synthesize contacts directly on the gripper's geometry, enabling zero-shot grasping of unseen objects. GOAG achieves an 86.93% average success rate on the Mu…
A developer who built a vulnerability triage CLI called triage-lens admitted they initially did not understand the EPSS, KEV, or CVSS scores the tool relies on. The tool, which passed 373 tests and shipped as v0.2.0, ran…
Grab, Southeast Asia's leading superapp, is hiring a Senior Data Scientist for its Bangalore, India office to build scalable AI solutions for GrabFin, its FinTech business spanning payments, lending, and insurance across…
Researchers from the University of Modena and Reggio Emilia's AImageLab introduced ReT-2, a unified multimodal retrieval model supporting queries and documents composed of both images and text. ReT-2 uses a recurrent Tra…
A developer discusses how artificial intelligence is reshaping software quality assurance, highlighting AI-driven self-healing automation, test case generation, predictive defect analysis, and strategies for testing AI m…
A developer analyzed approximately 1,000 keywords using Python, LangChain, and LLM APIs to investigate the relationship between Google AI Overviews and organic search rankings. The research, published on heysho.com, foun…
A new paper submitted to arXiv on August 27, 2026, argues that the Gaussian kernel, also known as the squared exponential or radial basis function kernel, should never be used as a default in Gaussian process regression …
Security researcher Jordy Zomer released Lemmalog, a Datalog engine that lets AI agents maintain and retract facts with provenance, on August 28th. The open-source tool, published under the MIT license, separates fuzzy L…
Spotify engineer Erik Bernhardsson argues in a 2016 blog post that GPU-accelerated deep learning has a fundamental scalability problem: minibatch complexity scales linearly with the number of parameters, making GPUs inef…
A developer's hand-written Gemma 4 port in pure JAX runs across TPU v5e/v6e and NVIDIA T4G GPUs, revealing that while most of the code is portable, two hardware-specific issues break the abstraction. The heterogeneous he…
Researchers at the National Laboratory of the Rockies (NLR) have developed IPC-Fusion, an open-source toolkit that integrates data from cameras, radar, lidar, and connected vehicles to create a digital twin of intersecti…
Planet Labs, a satellite imaging company, is hiring a Senior Machine Learning Engineer in Arlington, VA, offering $161k–201k/yr, to develop novel embeddings-based change detection and computer vision techniques for Defen…