Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers are advancing autonomous coordination by using self-supervised goal-reaching, where AI agents learn to reach target states without complex reward systems. This method outperforms traditional multi-agent reinf…
Researchers have developed DrugAgent, a multi-agent system using large language models to assess drug-target interactions, achieving 98.8% faithfulness to input evidence on kinase screening data. The system integrates ma…
Researchers introduced TREK (Teacher-Routed Exploration via Forward KL), a new AI training method that enhances learning through unconventional exploration strategies. TREK significantly improved performance on mathemati…
Researchers introduced TREK, a method that improves AI problem-solving by using verified output trajectories to extend model learning. TREK boosted Qwen3-8B's performance on AIME 2024 from 36.9 to 40.3 and on AIME 2025 f…
Kyutai and the Mirelo team released MuScriptor, an open-weight decoder-only Transformer for multi-instrument music transcription to MIDI. The model, trained on synthetic and real recordings across three stages, achieves …
A developer demonstrates that the Kalman filter, a sixty-year-old control theory algorithm, functions as a memory system by continuously blending prior beliefs with new noisy observations. The filter's Kalman gain dynami…
A developer reports issues with their LSTM implementation in C++, noting that training error initializes around 4.5 but can spike into the 20s, which they have not seen with other networks. They suspect a matrix indexing…
MarkTechPost published a tutorial on building an autonomous data science agent using DeepAnalyze-8B, sandboxed code execution, and iterative analysis. The agent runs on a T4 GPU in Colab, processes e-commerce data, and g…
New research challenges assumptions about infinite-width neural networks, introducing a width-solid learnability theorem that shows infinite width does not guarantee an edge over polynomial width for learning specific ta…
A developer shared a pattern for integrating AI decision-making into Python automation pipelines, using a small AI prompt to classify files before routing them to deterministic scripts. The approach adds a triage step th…
Researchers have introduced a new approach to Markov decision processes (MDPs) called minimum action distance (MAD), which measures the minimum number of actions needed to transition between states. This self-supervised …
Researchers have developed physics-informed neural networks (PINNs) that embed the axisymmetric equations of linear elasticity to model wave propagation in bimaterial systems like steel and aluminum. The approach accurat…
Researchers developed PRML2, a hybrid framework combining Kalman filtering with machine learning, to improve vehicle localization accuracy in autonomous driving, especially when GPS signals are weak. The system uses phys…
Supervised Reward Inference (SRI) offers a new approach to AI goal understanding by mapping observed behaviors to rewards, outperforming traditional methods in flexibility and accuracy. SRI achieves near-ceiling performa…
Quantum kernels could revolutionize machine learning by encoding data into quantum states, but they face design challenges like exponential concentration. A new approach called Local-Global quantum kernels aims to addres…
New research reveals that the L0 hyperparameter in sparse autoencoders (SAEs) is critical for feature disentanglement in large language models, with improper tuning leading to mixed or degenerate features. The study intr…
Researchers developed a robotic throwing method using potential field state representation that achieves up to 90% success rates in cluttered environments with unseen objects. The approach, trained via kinesthetic demons…
Researchers developed a controllability-observability framework to reduce hidden-state redundancy in deep neural networks, achieving up to 72.95% state compression on MNIST with minimal accuracy loss and 70.94% compressi…
Researchers developed a deep reinforcement learning framework based on Proximal Policy Optimization to optimize battery charging for autonomous mobile robots in warehouses. The model dynamically selects charging stations…
Scientists used machine learning to analyze strain data from California's San Andreas Fault, discovering hidden slow-slip events that may influence low-frequency earthquakes. The findings bring researchers closer to unde…