Search signals, briefings, company results, benchmarks and glossary terms.
Search signals, briefings, company results, benchmarks and glossary terms.
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.
DORA is an asynchronous reinforcement learning system that speeds up LLM post-training by overlapping generation with model training, addressing the rollout phase bottleneck.
FinBench proposes a new benchmark for evaluating agentic financial forecasting models, focusing on time-gated calibration and uncertainty quantification.
Research proposes an end-to-end pipeline using LightGBM and TreeSHAP for explainable money mule detection from transaction and behavioral data.
New research introduces Fuzzy Spectral Region Decomposition (fSRD) for automated, interpretable modeling of complex nonlinear dynamical systems.
Research introduces SelectInfer, a method for selective neuron loading and computation to optimize LLMs for resource-constrained edge devices.
Research explores LLM-based theorem proving by operating on abstract syntax trees to reduce token consumption and API costs for program verification.
Research explores Monte Carlo Dropout and entropy-thresholded selective prediction for robust brain tumor MRI triage, focusing on model confidence.
Sobek proposes a new method for streaming equivariant tensor product convolutions, reducing memory and improving GPU utilization for graph neural networks.
Research explores INT8 quantization, channel pruning, and spiking neural networks to improve efficiency of EEG seizure detection.
Research identifies split conformal prediction maintains marginal coverage but fails silently on per-class coverage under distribution shift.
Research introduces a modular framework for verifying multi-agent system transitions and trajectories, operationalizing action language semantics.
Research proposes a TabTransformer augmented by Boundary-Seeking GANs for robust intrusion detection, addressing class imbalance and adversarial attacks.
Research introduces AEVAL, a deterministic testing framework for LLM agentic skill workflows to improve reproducibility and comparability.
A systematic review of Lipschitz continuity in deep learning explores its role in model robustness, generalization, and optimization dynamics.
Research explores self-modifying agents for formal mathematical reasoning, focusing on evolving proof workflows in Lean using compiler feedback and self-repair.
Research explores value-aware prediction for multi-agent systems to maintain coordination and operation during intermittent communication failures.
Researchers introduced a chemical validity operator to filter implausible AI-generated compositions, enhancing reliability in materials design.
A research paper proposes a predict-then-correct framework using few-shot continuous contextual bandits for adaptive retail demand forecasting.
Research proposes a method for efficiently evaluating new LLMs on question sets using historical data, constructing confidence sequences to track capability.
Research introduces SALT, a Salience-Aware Lexical Trie method for long-context compression addressing 'theme collapse' in LLM inference.
Research proposes Normalized Rewards to prevent over-optimization in Direct Alignment Algorithms (DAAs) like DPO, which can degrade LLM performance.
Researchers introduced ATLAS, a framework equipping Multimodal Large Language Models (MLLMs) with 'Think, Plan, Paint' reasoning for controlled image generation.
Research proposes a Novel Hybrid Quantum Reservoir Computing (nHQRC) framework to detect phase transitions in non-equilibrium dynamical systems.
Research proposes Precision-Varying Prediction (PVP) to robustify ASR systems against adversarial attacks by randomly sampling precision during inference.
Research explores Differentiable Logic Gate Networks (Diff-Logic) for low-latency EEG classification on edge devices, compiling models into Boolean circuits.
Research explores hardware-aware design and optimization for edge intelligence systems, addressing complexity of deep learning on heterogeneous edge devices.
Research indicates constant-stepsize SGD's behavior fundamentally changes for convex objectives with flat minima, deviating from Gaussian limits.
Research evaluates machine learning models for Type 2 diabetes risk prediction, emphasizing external validation and fairness analysis on national populations.
Research introduces a Stochastic Dimension Zeroth-Order Estimator to improve memory and training efficiency for Physics-Informed Neural Networks (PINNs).
A research paper proposes an 'AI epidemiology' framework to standardize expert-AI interaction data for prospective risk detection in deployed AI systems.
Researchers introduced a dataset and method for text-prompted, context-dependent human visual similarity judgments to improve perceptual metrics.
Research disentangles model (epistemic) and data (aleatoric) uncertainty in facial age estimation using Bayesian Neural Networks.
New benchmark, 'Adaptive Adversaries,' evaluates LLM agent security against multi-turn, adaptive prompt injection and manipulation attacks.
Research reveals gradient-free privacy leakage in federated language models (FLMs) through selective weight tampering, exposing sensitive client data.
ZifaMem proposes a structured memory system for AI companions, organizing dialogue into session summaries, episodic memories, and a consolidated user model.
Research introduces a method for decomposing uncertainty in Bayes-filtered transformers, separating aleatoric from epistemic uncertainty for better risk assessment.
Research challenges the 'edge-of-chaos' heuristic for reservoir computing, finding optimal forecasting performance doesn't align with chaotic boundaries.
Research proposes mHC-GNN, a Graph Neural Network architecture adapting Manifold-Constrained Hyper-Connections to mitigate over-smoothing and enhance expressiveness.
Research introduces a perturbation-based method for LLM extrapolation, reinterpreting models through pre-post-additive noise for next-token prediction.
Research on reinforcement learning controllers for multi-agent networks shows sustained non-stationarity damages internal state and learning capacity.
Research uses topological methods to understand TabPFN's reliability on complex tabular data, enhancing model interpretability.
A research paper proposes a volatility-aware machine learning approach to detect extreme price movements in high-frequency financial markets.
Research identifies and removes an artifact in Nash equilibrium selection by regularized solvers in zero-sum games, aligning empirical results with maximum-entropy selection.
Research explores mixed-timescale differential coding for federated learning model broadcast to reduce communication load in wireless environments.
Research paper benchmarks classical and quantum machine learning models for image recognition on MNIST dataset, exploring quantum as a new paradigm.
SEC-bench Pro is a new benchmark evaluating frontier LLMs' ability to find real software vulnerabilities and generate proof-of-concept exploits.
Research explores OrthoGrad, a geometric intervention on optimizer updates, to reduce neural network memorization of noisy labels in training.
Research explores transferable low-rank convolutional bases to onboard unseen medical imaging modalities without retraining the core model.
Research explores methods to improve LLM planning capabilities by addressing the gap between theoretical Turing-completeness and empirical performance.
Research introduces AOE, a method improving out-of-distribution detection for ML models by recalibrating outlier labels during training.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion