Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Google's Project Suncatcher will launch its first physical test payload, a refrigerator-sized Planet Labs satellite carrying four Google Trillium Tensor Processing Units, on a SpaceX Falcon 9 rideshare mission from Vande…
SlopTotal, a free self-hosted open-source AI text detector from developer pablocaeg, runs 23 independent detection engines in parallel on a user's own CPU and reports an overall AUC of 0.974 on a 110-text RAID corpus, wi…
Researchers from Johns Hopkins University found that lower bone mineral density in the thoracic vertebrae is associated with faster cognitive decline and greater accumulation of white matter hyperintensities in the corpu…
Researchers introduced VLA-Precision, a method using asymmetric co-bootstrapping for efficient real-world online reinforcement learning of vision-language-action models, addressing the unreliability of pretrained VLA mod…
Google DeepMind's WeatherNext Cyclones AI forecast model won the 2026 Gizmodo Science Fair after a Nature study found it delivers, on average, an extra day of predictive accuracy for tropical cyclone track, intensity, an…
Towards Data Science published a guide on detecting hidden data drift in machine learning features using adversarial validation with scikit-learn. The article addresses the case where individual feature distributions app…
Researchers propose enhancing photogrammetric Digital Surface Models (DSMs) derived from satellite stereo imagery by using pretrained diffusion models with multimodal conditioning, addressing noise, outliers, and voids t…
Idiap Research Institute researchers released a standalone demo of Hiera-Hand, a hand-action recognition pipeline that tracks people, localizes their hands, and labels every hand in every frame as grasp, hold, operate, o…
A teacher-guided low-fidelity fitness proxy combined with constrained evolutionary hardening and a trust-region sequential quadratic programming backbone can cut evaluation cost by more than 50% while preserving rank ord…
A developer reports that instructing a large language model to conduct its internal reasoning in Roman Urdu caused it to narrate the instruction in English rather than actually think in the requested language, while sust…
Zalando tested a daily machine-learning pricing system across 23 A/B tests in 12 European markets and found it improved profit during sales without reducing revenue or units sold after returns, according to a company-aut…
A developer modified the Qwen 2.5 Omni 3B model's attention mask to stop slide images from interfering with audio transcription when slide context is supplied alongside audio. The approach first ablates the visual span o…
A sparse-probing study on a frozen BERT-base-uncased encoder identified fewer than 1% of the model's 9,216 CLS hidden-state units (12 layers × 768) as stable AI-text detectors across six generators on the RAID benchmark,…
Pretraining small decoder-only transformers on code-switched text aligns representations of parallel text across English, Dutch, and Chinese and improves scores on the BabyLM evaluation suite, according to an arXiv paper…
Researchers introduced MemProbe, a cognitive-science-inspired framework for diagnosing stability-plasticity tradeoffs in agent memory, detailed in arXiv paper 2609.30558v1. MemProbe provides four reusable experimental pa…
A new arXiv paper (2609.30397v1) proposes a synthetic ground-truth framework for evaluating explainable AI (XAI) methods, using controlled interventions to build datasets where input-component importance is known by desi…
A new algorithm called Spectral Feedback improves test-time alignment of discrete diffusion models by selecting which token positions to re-mask and re-sample, according to an arXiv paper (arXiv:2609.30456v1). Applied to…
A new arXiv paper (2609.30484v1) proposes a knowledge graph-based evaluation framework for testing whether large language models genuinely comprehend context in question answering rather than relying on surface-level pat…
A new arXiv paper (2609.30469v1) reports that an external LLM-as-a-judge filtering paradigm significantly reduces word error rate in pseudo-labeled training sets for noisy Broadcast Police Communication audio from Baltim…
A paper posted to arXiv (2609.30550v1) introduces Benchy, a semantic language and execution engine that specifies an AI benchmark as a program, scoring function, and dataset, B=(P,S,D), kept separate from the AI system u…