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.
Research identifies "Unlearning Aftermaths" in Vision-Language Models (LVLMs) after privacy-driven unlearning, leading to degenerate or hallucinated outputs.
Research identifies a mechanistic explanation for catastrophic loss explosions during low-precision transformer training with Flash Attention.
Research introduces a linearity score (λ(f)) to diagnose neural network input-output behavior, claiming fidelity to models is insufficient for XAI.
Research explores active sequential prediction-powered mean estimation, deciding when to query ground-truth labels versus using model predictions.
Research claims ML-enhanced Monte Carlo outperforms classical methods for some Quadratic Unconstrained Binary Optimization (QUBO) problems.
SeekerGym is a new academic benchmark evaluating AI agents for reliable information seeking, focusing on completeness and bias in retrieval.
New benchmark, MMErroR, evaluates Vision-Language Models' ability to detect and categorize reasoning errors in multi-modal inputs.
Research identifies logit suppression vulnerabilities in LLM safety alignment, enabling manipulation despite current safeguards.
A research survey from arXiv explores methods to address distribution shifts in deep learning models for medical image analysis, enhancing deployment reliability.
Research proposes ASTRA, an automated framework to autonomously discover, retrieve, and evolve LLM jailbreak attack strategies through continuous learning.
CaTS-Bench introduces a new benchmark for evaluating language models' ability to describe time series data across 11 diverse domains.
Research evaluates LLM alignment with human moral values in high-stakes kidney allocation, identifying deviations from human preferences.
Research paper surveying Bayesian Neural Networks, a method to quantify predictive uncertainty in deep learning models.
Research claims simplified optimizers during LLM unlearning improve the robustness of unlearning effects, making them less susceptible to post-processing neutralization.
Research paper explores using differential privacy techniques to mitigate overfitting in deep neural networks, improving model generalization.
Researchers propose D-QRELO, a training- and data-free delta compression method for fine-tuned LLMs, addressing memory overhead for large SFT datasets.
Research investigates how defensive training methods like Positive Preventative Steering (PPS) and Inoculation Prompting (IP) protect LLM integrity.
SLO-Guard is a crash-aware autotuner for vLLM serving that optimizes LLM inference under latency SLOs while managing budget constraints.
Research surveys Graph Neural Network (GNN) architectures designed for heterophilous graphs, where connected nodes often have different labels.
Alibaba's AliExpress developed SIGMA, a generative multi-task recommender using LLMs for semantic-grounded, instruction-driven recommendations.
New research introduces TransXion, a high-fidelity graph benchmark designed to improve anti-money laundering (AML) machine learning models by addressing limitations in existing datasets.
Research claims safety alignment in LLMs erodes during continual domain adaptation, addressable by SafeAnchor to prevent cumulative safety failures.
Research explores in-context learning's robustness in non-stationary environments, critical for time-series forecasting and control with foundation models.
Research evaluates how using off-policy or synthetic LLM responses for training probes impacts their ability to detect concerning behaviors.
Research identifies 'Visual Dominance Hallucination' in MLLMs, where imperceptible visual changes bypass price constraints in financial transaction agents.
Research paper proposes a homomorphic encryption (HE) method for low-latency, memory-efficient, high-throughput batch inference on encrypted neural networks.
Research identifies a "Scaling Law of Miscalibration" in on-policy distillation (OPD): models show improved accuracy but severe overconfidence.
RAYEN framework enforces hard convex constraints on neural network outputs, guaranteeing satisfaction during training and inference.
Research proposes a scalable Nystrom-based kernel two-sample test with permutations, enhancing Maximum Mean Discrepancy (MMD) for large datasets.
Research paper introduces subbagging and adaptive cross-bagging to improve random seed stability and reproducibility in ML-based estimation.
A study found security training improved security quality in LLM-assisted Java Spring Boot backend development among 12 developers.
SafeLM proposes a federated learning framework integrating gradient smartification and Paillier encryption to address LLM privacy, security, and robustness.
Research identifies 'Format-Reliability Gap' where LLMs generate insecure code but can identify/explain the vulnerability when prompted directly.
Researchers propose a parallel training framework for Graph Transformers, addressing single-GPU limitations and out-of-memory issues on large graphs.
Research proposes a framework to infer sensitive attributes from auxiliary features to enforce fairness constraints in high-dimensional generalized linear models.
FairLogue toolkit evaluated intersectional fairness in clinical ML models using the All of Us dataset, revealing compound disparities.
Research identifies high-gradient samples during fine-tuning as primary cause of large language model safety alignment drift, impacting refusal and truthfulness.
Research explores LLM reasoning improvements with weak supervision for reinforcement learning (RLVR), addressing challenges in reward signal construction.
Research proposes a two-rate error measurement for LLM protocols to audit correction vs. corruption, improving understanding of their impact.
Research proposes a machine learning approach to solve two-stage adaptive robust optimization problems with binary here-and-now variables.
Research proposes E-Value based stopping rules to make Bayesian Deep Ensembles (BDEs) more computationally efficient for uncertainty quantification.
Research highlights misalignment between LLM benchmark performance and actual downstream impact, especially in difficult-to-verify tasks.
Research predicts LLM compression degradation using spectral statistics across Qwen3 and Gemma3, avoiding costly full model evaluations.
Research claims Reinforcement Learning with Verifiable Rewards (RLVR) can be effective for fine-tuning LLMs with limited data and compute.
Research surveys streaming LLM architectures for dynamic, real-time scenarios, aiming to clarify fragmented definitions and taxonomies.
BenchMarker, an LLM-powered toolkit, identifies contamination, shortcuts, and writing errors in multiple-choice NLP benchmarks using an education rubric.
Researchers introduced LiveFact, a dynamic, continuously updated benchmark designed to evaluate LLM performance on time-aware fake news detection.
Research introduces BASIL, a new Bayesian method to detect and measure sycophancy in LLMs, distinguishing it from rational behavior shifts.
Research identifies 'explanation bias' in post-hoc feature attribution methods, showing varied token-level insights due to lexical and position preferences.
NL2SQLBench introduces a modular framework to evaluate large language model-enabled Natural Language to SQL solutions, addressing a gap in systematic LLM NL2SQL benchmarking.
© 2026 OneBench: AI Insights. All rights reserved.
Evidence before opinion