Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers at an undisclosed institution applied Group Relative Policy Optimization (GRPO) to adapt an LLM-based automatic speech recognition (ASR) system to synthetic speech, achieving a 40% relative word error rate (W…
A new review of remote sensing foundation models (RSFMs) finds that no single geospatial foundation model is universally best and that inconsistent evaluation hinders fair comparison and reliable deployment. The chapter …
Researchers identified failure modes in Hamiltonian Generative Networks that prevent temporal generalization in non-conservative environments, and proposed targeted fixes enabling stable dynamics prediction at variable t…
Researchers introduced AgentNAS, a method that uses a large language model to generate a seed architecture and decompose it into a slotted architecture for conventional neural architecture search, achieving state-of-the-…
A study comparing LSTM networks with traditional machine learning models for sentiment analysis on Twitter data found that LSTM outperformed logistic regression, random forest, naive Bayes, and gradient boosting, achievi…
Researchers propose SAGA, a training-free stable acceleration guidance method for autoregressive video generation that reduces temporal errors like flickering and motion jitter by using spectral guidance and noise initia…
Researchers found that best-of-N text-to-speech evaluation is confounded by alignment between ASR verifier and evaluator families, with same-family pairs recovering 2-3× more oracle headroom than cross-family pairs. They…
Researchers introduced Omni-Sleep, a foundation model for sleep analysis that leverages the physiological hierarchy of the central and autonomic nervous systems to learn structured representations from polysomnography da…
Researchers introduced FedTR, a federated learning framework incorporating transfer learning for industrial visual inspection, achieving 95.5% word-level accuracy on homogeneous data and 94.2% on heterogeneous data for l…
A new survey highlights how combinatorial optimization can enhance trustworthiness in machine learning by providing global guarantees, formal certificates, and explicit trade-off handling across tasks like interpretabili…
Researchers introduced DreamCharacter-1, a lightweight post-adaptation framework that enhances pretrained 3D foundation models for high-fidelity, production-ready 3D character generation. The framework includes geometry …
Researchers introduced Meta Neural Cellular Automata (MetaNCA), a framework that learns local rules to self-organize the weights of artificial neural networks without backpropagation. The method generates weights for div…
Researchers introduced Lipschitz Scaling Training (LiST), a method that adjusts the Lipschitz constant of neural networks to achieve out-of-the-box calibration while balancing accuracy and robustness. LiST outperforms ba…
Researchers propose a three-phase pipeline to improve small language models for low-resource programming languages like Julia and Ballerina, achieving up to +14.2 points on pass@1 metrics while reducing data and cost by …
Researchers propose a hybrid classical-quantum image classification framework using a mixture of experts, achieving a two-fold reduction in failure rate on MNIST and Fashion-MNIST benchmarks compared to individual expert…
Researchers proposed a structured pruning method for large language models that combines power transformation, sign-preserving score aggregation, and adaptive feature retention. Tested on Llama-3-8B, Vicuna-v1.5-13B, and…
Researchers propose LOGOS, a transformer-based model that uses textual prompts to guide oriented object detection in aerial and satellite imagery. The approach outperforms existing methods on the DOTA dataset, particular…
A new survey from arXiv explores post-training techniques for end-to-end autonomous driving models, addressing limitations of open-loop imitation learning in safety-critical environments. The paper categorizes existing m…
Researchers introduced LEXIC, a lightweight eye-tracking extension that injects word-level difficulty signals into gaze-only models, achieving AUROC gains of up to +2.9 percentage points on reading comprehension tasks wi…
Researchers introduced JAM, a theory-agnostic framework for personality recognition that uses an LLM-as-a-Judge mechanism to identify ambiguous samples and guide adaptive metric learning, improving cross-framework genera…