Use this view to inspect the underlying evidence corpus. For ranked developments, decision posture and interpretation, use Signals.
Raw feed or Signals?
Raw feed is chronological evidence. Signals ranks and interprets material change.
Researchers introduced the Tversky Monosemanticity Score, a new metric for evaluating monosemanticity in Sparse Autoencoders to improve explainability without external labels.
Researchers propose ETAS, a new programming language designed for agent systems that semantically models agents, tools, prompts, and memory.
Research investigates if adding a reasoning step to Vision-Language-Action (VLA) models improves robustness against perturbations at different stages.
Research explores how entanglement geometry in quantum circuits affects the feasibility of circuit cutting, classical hardness, and trainability.
Research challenges the 'edge-of-chaos' heuristic for reservoir computing, finding optimal forecasting performance doesn't align with chaotic boundaries.
Researchers introduced a chemical validity operator to filter implausible AI-generated compositions, enhancing reliability in materials design.
Research proposes AutoEncoder-Compressed Parallel Split Learning to reduce communication overhead and compute for fine-tuning large models on edge devices.
Research explores value-aware prediction for multi-agent systems to maintain coordination and operation during intermittent communication failures.
Research on reinforcement learning controllers for multi-agent networks shows sustained non-stationarity damages internal state and learning capacity.
Research explores multi-agent policy optimization for cooperative Multi-Agent Reinforcement Learning (MARL), focusing on optimal agent aggregation.
Research proposes hardware mechanisms to dynamically throttle AI performance, offering a last line of defense against potentially unsafe AI intent.
SciForma is a research paper exploring structure-faithful generation of scientific diagrams, emphasizing accuracy in components and relations.
New research proposes a covariance-induced fairness gap penalty to improve subgroup-fair clustering, addressing computational and stability issues with multiple sensitive attributes.
ClouDens is a research paper on operational context-aware anomaly detection for large-scale cloud system monitoring using multivariate time series data.
New research explores certified training to improve the formal robustness guarantees of vision models against runtime perturbations like motion blur.
Research introduces ATLAS, a multi-environment factor model to identify invariant and transferable latent factors across heterogeneous data environments.
Research proposes RAG-based policy learning using vector search for action selection in causal inference, estimating conditional outcomes.
DARPA's AlphaDogfight Trials leveraged hierarchical reinforcement learning for autonomous air combat, addressing high-dimensional control challenges.
Research details the existence and structure of infinite-dimensional null spaces in two-layer neural networks, identifying parameter nonuniqueness.
Research paper overviews Automated Reinforcement Learning (AutoRL) methods for optimizing RL problem modeling, algorithm selection, and hyper-parameters.
Research surveys methods for representing black-box single-objective continuous optimization problems using features for ML tasks like algorithm selection.
Research on regret optimality of Sample Average Approximation (SAA) for general data-driven newsvendor problems, moving beyond linear costs.
Research explores online-score-aided federated learning for resource-constrained wireless clients with continuous data streams.
Researchers developed an attention-based deep neural network using cloud images to improve short-term distributed solar generation forecasting.
Research identifies backdoor vulnerabilities in third-party task vectors, allowing malicious actors to embed hidden threats in modular model updates.
A research paper proposes PLayer-FL, a principled federated learning approach to address non-IID data performance degradation via layer-wise federation.
Research introduces AdaGC, an adaptive gradient clipping method to enhance the stability of large language model pretraining by mitigating loss spikes.
Research on Supervised Reward Inference (SRI) proposes a method to infer human goals from suboptimal or communicative behaviors, improving reward model robustness.
A research survey on "unlearnable data" (ULD) explores methods to prevent ML models from learning patterns for privacy and security through data perturbations.
Research identifies specific reasons why physics-informed machine learning (PIML) fails in traffic flow modeling, underperforming data-driven and physics-based baselines.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion