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 explored diffusion generative models to address limited training data for Auditory Attention Decoding (AAD) using EEG in hearing aids.
Research proposes Decode-Time Grammars for constrained LLM generation, improving code accuracy for domain-specific languages and APIs.
Research explores decentralized multi-agent reinforcement learning (MARL) for improving resilience in critical infrastructures.
A new benchmark, HALLMARK, diagnoses three failure modes in LLM citation verifiers, addressing high rates of hallucinated references in academic papers.
Research compares TF-IDF + Logistic Regression and fine-tuned DistilBERT for phishing detection against adversarial evasions.
Research explores using binary silver labels, like diagnosis codes or NLP mentions, to improve weakly supervised phenotyping algorithms in EHR data.
Recti-Q improves out-of-distribution robustness for quantized vision models, addressing reliability degradation in edge robotic perception.
Research details Quantum Reservoir Computing (QRC), using quantum system dynamics for temporal input transformation and classical readout training.
Research introduces a method for exact recovery of Gaussian graphical models from a single trajectory of Gaussian Glauber dynamics.
Research identifies a new inference-time attack surface against graph foundation models via their shared representation ('alignment layer').
Research introduces E2E-CDiff, a conditional diffusion model generating realistic, controllable visual traffic scenarios for autonomous driving evaluation.
Research addresses AI model generalization failure in breast MRI tumor classification across institutions due to domain shift and dataset bias.
Research explores using algebraic signatures derived from observed vanishing binomials in probability tensors to identify probabilistic model structure.
Research introduces LLIFT, a method using GAN- and diffusion-based approaches for generating ground-truth medical images with known features.
Research benchmarks deep learning models for layout detection and information extraction specifically on Architecture, Engineering, and Construction (AEC) drawings.
Research details the exact error conditions required for tractable, unbiased sampling with inexact score oracles, ruling out weaker error assumptions.
Research introduces boundary-enforced Physics-Informed Neural Networks (PINNs) with a priori error bounds for elliptic Dirichlet problems.
ABot-World-0, an action-conditioned video world model, enables real-time, long-horizon interactive world rollout on a single desktop GPU.
Research introduces ResearchArena, a framework to evaluate AI agent sabotage and monitoring in automated AI R&D, focusing on safety assessment.
Research explores fundamental performance limits of constructing K-class classifiers from O(log K) simple binary classifiers in a distributed manner.
Research introduces 1-Lipschitz neural networks on Hadamard manifolds, aiming to improve robustness and stability beyond Euclidean space methods.
Soft-TransFormers (Soft-TF) introduces a continual learning method for pre-trained Transformers using task-specific soft subnetworks.
A research survey explores Parameter-Efficient Continual Fine-Tuning (PECFT) methods for large pre-trained models to adapt to dynamic data.
Research explores Topology-Driven Clustering using Betti Number Filtration to enhance performance for datasets with complex geometric structures.
New research proposes Chi-Square Wavelet Graph Neural Networks for anomaly detection in complex, heterogeneous networks, addressing GNN limitations.
Research finds that methods to accelerate LLM evaluation by shrinking benchmark datasets often miss the mark, systematically assessing 11 methods.
Research explores federated learning (FL) for continuous client onboarding without catastrophic forgetting, using hypernetworks and data-free replay.
Research paper HERAKLES explores hierarchical skill compilation for LLM agents in open-ended, partially observable, sparse-reward environments.
Research proposes Non-Trainable Modification (NTM) neural network architecture to compute Nash equilibria in stochastic differential games on graphs.
Research details BETA, an indirect evasion attack targeting Graph Neural Network-based anomaly detection in sensor networks by perturbing non-target sensors.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion