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.
A research paper demonstrates standard conformal prediction fails in minority class coverage under severe imbalance, proposing cost-sensitive CP.
Academic research demonstrates that accurate ML-based option pricing does not guarantee accurate recovery of latent risk-neutral densities.
Academic researchers expose a critical vulnerability where poisoned ML model backdoors trigger even under unseen inference variations.
Research demonstrates that simple blind resampling outperforms complex self-repair feedback loops for small code generation models.
Researchers identify an exponential 'incast' network bottleneck in MoE model routing and propose a proactive rate-based scheduling fix.
Researchers identify double descent phenomena in sparse sensing and reduced-order models, offering mitigation strategies for noisy reconstructions.
Researchers propose LADDER, a domain generalization method using reweighting instead of strict invariance to handle varying causal mechanisms.
Researchers propose a system using RAG and agentic LLM personas modeled on individual FOMC members to predict Federal Reserve rate decisions.
Research explores stable and budget-feasible coalition formation for clustered federated learning, using a hedonic potential-game approach.
New research proposes a probabilistic K-line forecasting method that ensures consistency by preventing quantile and K-line crossing without retraining.
ToxScreen, a new research technique, detects backdoors from poisoned training data in LLMs, even without original data or trusted models.
Research explores using graph-based inference to identify business conduct risk when reported incident data is sparse and biased, acknowledging silent risk.
Researchers propose FedDAB, a two-phase defense method using local contrastive regularization and alignment checking to counter backdoor attacks in Federated Learning (FL).
Research introduces 'feature instability' (FI) to measure the stability of feature bagging, an ensemble method for machine learning.
Research details an algebraic framework to understand the computational limits of causally masked transformers under finite precision and real-world constraints.
Researchers introduced HoF-Bench, a new benchmark of 95 real AI-discovered CVEs, to evaluate LLM-based security analyzers like AISLE's.
Research addresses compounding error in video diffusion models for long-horizon generation, proposing a regularization technique to improve frame quality.
GPTQ-2D introduces a two-sided adaptive rounding method for quantizing large language models, improving on prior quadratic metric-based approaches.
Researchers propose Physics-Informed Kernel Methods (PIKS) as an alternative to PINNs, offering closed-form solutions for integrating physical principles into ML models.
InferScale proposes GPU-native KV injection to optimize LLM serving for personalized contexts, avoiding repeated prefill of long conversation histories.
Researchers introduced a novel statistical framework for evaluating bias in medical imaging machine learning models, using counterfactual analysis to assess dependency on sensitive attributes.
Research paper explores low-precision training techniques for large language models to improve efficiency and scalability by reducing hardware resource requirements.
Research identifies dense local dependencies as a key cause of attention-logit explosion and training instability in long-sequence transformers.
Research on arXiv explores fine-grained training methods, demonstrating enhanced performance in classification tasks by utilizing hierarchical labels.
Researchers propose PrunedLoRA, a method leveraging structured pruning to create more expressive and compact low-rank adaptations (LoRA) for fine-tuning LLMs.
Research proposes a structural separation principle for predictive uncertainty in latent variable models to reduce correlation between epistemic and aleatoric estimates.
Researchers introduced MapTab, a diagnostic benchmark for evaluating Multimodal Large Language Models (MLLMs) on long-horizon, multi-criteria reasoning tasks using topological graphs.
Research demonstrates transformers can learn and apply novel rules beyond their training data through computation, not just interpolation.
Research demonstrates a method for exact symmetry enforcement in machine learning via tensor calculus, eliminating compounding equivariance errors.
Research proposes a novel method to address ambiguity in dimensionality reduction visualizations by splitting ambiguous data points for clearer insights.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion