Inspect benchmark version, contamination controls, sample size, held-out tasks, adversarial failures and scoring rules. Higher scores alone do not establish lower production risk.
Evaluation & safety · primary_technical_publication
Evaluating Microsoft Copilot at enterprise scale requires test data that captures the full complexity of real organizations, yet real enterprise data contains sensitive information that
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · preprint
Vision-language models (VLMs) exhibit strong zero-shot generalization but remain vulnerable to adversarial attacks. Existing adversarial fine-tuning methods often use pre-trained model guidance to preserve image--text
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · preprint
Reinforcement learning (RL) is increasingly used to train robust, adaptable systems across healthcare, e-commerce, and robotics. In these settings, legal, ethical, or logistical concerns may
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · preprint
Mobile money (MoMo) is Ghana's preferred retail payment channel, recording GHC518.4 billion in transactions in December 2025 and over 26.7 million active accounts (Graphic Online,
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · preprint
How effectively can vision-language-action (VLA) models learn diverse language-conditioned behaviors as training data scales? Investigating this question requires large and varied robot datasets, whose collection
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · preprint
Monocular RGB tracking is useful for robots operating near people, but perception errors affect binary stop/proceed decisions in ways that geometric accuracy metrics do not
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · preprint
Choosing a robot action model requires evidence about task capability, adaptation cost, deployment requirements, and the reliability of reported conclusions. Success rates alone offer limited
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · primary_technical_publication
Generating tokens faster starts with identifying what limits inference performance. In the batch-size-1 evaluation of Ling-3.0-flash from Inclusion AI on NVIDIA...
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · provider_announcement
OpenAI and Poland’s NASK are partnering on AI safety, security, and technical research to strengthen safeguards and support a more resilient digital ecosystem.
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · provider_announcement
OpenAI outlines a blueprint for empowering cyber defenders with frontier AI, stronger safeguards, modernized cybersecurity, and more resilient public institutions.
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · provider_announcement
Learn how OpenAI is improving ChatGPT across low-resource programming languages, with new evaluations for software engineering, SQLite, and COBOL migration.
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Evaluation & safety · provider_announcement
Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
Automatisch gesichtet · hier nicht fachlich begutachtetForschungsnotiz lesen ↗Directly relevant to building realistic test tasks without exporting private Sevensa or customer records.
Source: Microsoft Research · 2026-10-30
Separate AI source check · not peer review