Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers introduced PhoneHarness, a mixed-action benchmark and execution harness for phone-use agents that combines GUI, CLI, and tool actions. The system achieved a 75.0% pass rate on verifiable mobile workflows, out…
NVIDIA released Nemotron 3 Ultra, a 550B-parameter hybrid Mamba-Transformer model with 55B active parameters, achieving up to 6x higher inference throughput than state-of-the-art LLMs while maintaining accuracy. The mode…
Researchers achieved 98.91% accuracy in multimodal emotion recognition from physiological signals by combining LSTM, TCN, and Transformer models with late-fusion ensemble on the WESAD dataset. Transformer models excelled…
Researchers introduced a practical evaluation method for long-form simultaneous speech-to-speech translation (SimulS2ST), using ASR and forced alignment to compute sentence-level latency and quality metrics. Experiments …
Researchers propose QPILOTS, a method that steers flow-matching and diffusion policies at inference time by projecting noisy intermediate states to clean action estimates for critic gradient computation. QPILOTS achieves…
Researchers introduced GRAPE, a training framework that progressively exposes neural network parameters to improve adversarial robustness. On CIFAR-10 with ResNet-18, GRAPE achieved 56.94% PGD-20 robust accuracy, a 5.24 …
Researchers propose GRASP, a method for multi-source transfer learning that achieves 93.5% mean accuracy across benchmarks while using constant memory, compared to 71.7% for ensemble methods. GRASP sequentially merges so…
Researchers formalized embedding model routing as an adversarial contextual linear bandit with low-rank experts, identifying a log-quadratic policy class for efficient online learning. They proposed the Hypentropy Policy…
Researchers introduced the Separable Neural Architecture (SNA), a function representational class combining neural approximation with tensor decomposition, which decouples localized coordinate functions from global inter…
A new study compares 19 graph neural network layer types for driving trajectory prediction, finding that ARMA, Chebyshev, and topology-aware layers outperform others. The research provides design principles such as sum-b…
Researchers used machine learning models to predict exam outcomes from physiological data such as electrodermal activity, heart rate, and skin temperature. Deep learning models like transformers and LSTMs captured comple…
Researchers introduced FASTMIX, a framework that automates data mixture optimization for training large models by reformulating mixture selection as a bilevel optimization problem and using gradient descent to jointly op…
Researchers trained a two-layer transformer encoder to classify rational elliptic curves by rank, achieving over 99% accuracy. The model learned the Mestre-Nagao heuristic from data alone, with input weights matching the…
Researchers introduced a temporal difference (TD) objective for diffusion models that enforces cross-time consistency along the denoising trajectory, improving sample quality especially with few sampling steps. The metho…
Researchers introduced Physics-conforming Latent Twins, a framework for learning surrogate models that respect physical laws like conservation and dissipation. The method constrains latent dynamics to preserve structural…
Researchers propose replacing the inner product score in sparse autoencoders with a learned blend of cosine similarity and input magnitude, finding that cosine-scored SAEs learn more human-recognizable features and avoid…
A study from arXiv finds that linear regression outperforms the naive random walk in forecasting the monthly USD/CAD exchange rate, while machine learning ensemble models show only marginal differences. The research, usi…
Researchers proposed a sub-quadratic vision transformer for image captioning that replaces standard self-attention with a Gaussian Mixture Model-based clustering mechanism, reducing computational complexity from O(n^2) t…
Researchers developed a model and benchmark for attribute inference from interactive targeted ads, showing that repeated campaigns with identity exposure produce measurable but bounded inference signals, with disclosure …
Researchers introduced AdaNAGED, a zero-order parameter-free optimization method for LMO-based fine-tuning of large language models, addressing memory overhead and sensitivity to hyperparameters. The method unifies gradi…