Find a story
Search Spins
Search titles, summaries, and missing voices across published articles — press releases, announcements, and media coverage.
10 results for “robustness”
Demystifying Entropy-based Selection for Chain-of-Thought Compression in Large Reasoning Models
A new arXiv preprint challenges the efficacy of entropy-based pruning for Chain-of-Thought compression, finding no advantage over random pruning across models and tasks, and showing token-level entropy selection works only on math benchmarks due to numeric token properties—not generalizable reasoning heuristics.
Aug 3, 2026
Imbalanced Data Clustering via Targeted Data Augmentation Using GMM and LLM
A new unsupervised data augmentation method combining Gaussian Mixture Models and Large Language Models is proposed to improve clustering of underrepresented topics in imbalanced NLP datasets.
Aug 3, 2026
OpenAI details GPT-Red, an internal automated red-teaming model that scales prompt injection vulnerability discovery so it can fix bugs before wider deployment (OpenAI)
OpenAI announced GPT-Red, an internal AI model designed to automatically detect prompt injection vulnerabilities in its systems before public deployment, framing it as a proactive safety measure.
Jul 16, 2026
GPT-Red: Unlocking Self-Improvement for Robustness - OpenAI
OpenAI announced GPT-Red, a new AI system framework claiming to enable self-improving robustness through recursive red-teaming, though no technical details, empirical validation, or release timeline were provided.
Jul 16, 2026
GPT-Red: Unlocking Self-Improvement for Robustness
OpenAI announced GPT-Red, an internal automated red teaming system using self-play to test and improve AI model robustness against prompt injection and alignment failures.
Jul 15, 2026
Interval Certifications for Multilayered Perceptrons via Lattice Traversal
A new theoretical framework reduces adversarial robustness certification for multilayered perceptrons to a lattice traversal problem, introducing formally guaranteed sound and complete interval certifications with provable complexity asymmetries.
Jul 13, 2026
SHIFT: Survival Prediction from Incomplete and Heterogeneous Genomic Data
Researchers introduced SHIFT, a transformer-based survival prediction model designed to handle structurally missing genomic features across institutions without imputation, improving generalization in multi-center precision oncology.
Jul 10, 2026
Exogenous Dropout: A Simple, Strong Baseline for Corruption-Robust Time Series Forecasting with Covariates
A new training technique called 'exogenous dropout' improves robustness of time series forecasting models to corrupted or missing exogenous covariates without sacrificing clean-data accuracy, and is released as an open benchmark and baseline recommendation.
Jul 9, 2026
From Approximation to Emergence: A Theory of Deep Learning
A new arXiv monograph proposes a unified theoretical framework for deep learning, positioning emergence—not just approximation—as the central organizing principle of modern AI theory.
Published Jul 3, 2026 · Analyzed Jul 6, 2026
Learning dynamical systems from noisy data with Weak-form Kernel Ridge Regression
Researchers propose a new method for learning dynamical systems from noisy data.
Published Jul 2, 2026 · Analyzed Jul 5, 2026