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 propose DCS, a unified framework for detecting cross-modal copyright infringement in foundation model outputs by analyzing conditional sensitivity.
Researchers introduced CEL, a new library and benchmark for evaluating counterfactual explanations in explainable AI, focusing on properties like actionability.
Research proposes patient-aware sampling for pretraining EHR foundation models to address bias from mixed patient data and imbalanced contributions.
Research introduces 'Unbiased Open World Regularization' to address bias in self-supervised learning models and Joint-Embedding Predictive Architectures (JEPAs).
Researchers propose Class-Balanced Softmax, a Bayes Theory-based method, to improve deep learning model performance on imbalanced datasets, addressing limitations of existing methods.
Research introduces a framework for autoregressive EHR foundation models to incorporate multimodal inputs like ECGs, X-rays, and clinical notes.
New research introduces "interior interpretability" to characterize how Transformer layers compose, using attention rollout for tabular data models.
Research explores identifiability conditions for controlled world models using Joint-Embedding Predictive Architectures (JEPAs) for robust planning.
Research introduces κ-LoRA, a method to selectively update LoRA matrices based on condition numbers, aiming to reduce fine-tuning computational cost.
Research introduces Susceptible Architectures (SUSA), a new reservoir-design principle for regime-conditioned volatility forecasting using complex-valued open-chain and periodic reservoirs.
New research proposes Spectral Flow Certificates to predict Graph Neural Network (GNN) performance on long-range tasks before training, based on graph structure.
Researchers introduced SCOPE, a benchmark for schema induction and fusion from text, and SCION, an auditable reference pipeline for the same task.
Research finds LoRA struggles to internalize multi-step procedural knowledge in large language models, failing to match full fine-tuning.
Researchers propose FrED, a black-box probabilistic framework for training data attribution in generative AI, fusing feature similarities with domain knowledge graphs.
Research identifies 'Role Drift' in compound LLM systems, where modules deviate from assigned roles while maintaining overall task performance.
Research proposes treating LLM prompts as a first-class data type within database management systems for better optimization and governance.
Research explores extending Lustre's clock calculus to machine learning dataflow networks for static property determination like liveness and memory bounds.
Research explores invariant measures to understand long-term statistical behavior of the Multiplicative Weights Update algorithm in chaotic game dynamics.
Research reveals audio deepfake detectors' error rates increase twentyfold on new datasets due to over-reliance on speaker identity in training.
Research studies the impact of AI coding agents on software development, specifically how their contributions evolve across the SDLC in pull requests.
Research identifies assumption gaps in privacy-preserving machine learning auditing protocols, impacting their real-world applicability for accuracy and fairness.
Research introduces Simulation-Based Empirical Bayes (SBEB) for simultaneous inference when likelihoods are available only through simulators.
Researchers propose a novel online watermark detection framework for LLM-generated text, enabling early stopping for efficiency in streaming content.
Research proposes a "Reasoning Denoiser" to improve hallucination detection in large reasoning models by identifying and removing irrelevant or repetitive steps in reasoning traces.
A research paper proposes Deep Convolutional Large-Margin $\ell_p$-SVDD for visual anomaly detection, combining deep features with explicit margin-aware boundaries.
Research identifies explicit iteration complexity for exact data-driven inverse optimization of Integer Linear Programs using gradient-based methods.
Research proposes General Value Functions (GVF) for remaining useful life (RUL) and failure-mode prediction in predictive maintenance.
Hopformer, a two-stage Transformer framework, is introduced for multi-variate time series forecasting by separating common trends from specific information.
Researchers propose SEM-DNN, a heteroscedastic neural simultaneous-equation estimator, to learn bidirectional causal interactions from observational data.
Research explores agentic AI for automated, evidence-grounded root cause analysis of industrial anomalies, addressing explainability and data scarcity.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion