NTT Scientists Contribute Fifteen Research Papers to NeurIPS 2025

What we drew from this source

The claims Via News extracted from this document. We point to the source; we don't replace it.

• AI is becoming ubiquitous, but how these computational engines actually work remains—to a surprising degree—a mystery, which is why our scientists keep probing with fundamental questions

  80% confidence

• Standard SAEs architectures embed implicit geometric assumptions that not only reveal concepts inside models but also shape what concepts can be discovered

  80% confidence

• With the EU AI Act mandating watermarking, the topic has become increasingly urgent, and even state-of-the-art watermarking schemes can be reverse engineered, revealing significant risks for enterprises

  80% confidence

• Researchers must keep pace with operational challenges in AI development

  80% confidence

• Understanding training dynamics in in-context learning may enable engineering reliable ICL behavior

  80% confidence

• The watermarking vulnerability findings expose a foundational vulnerability in today's AI trust and traceability mechanisms

  80% confidence

Cited in these Via News reports

• Enterprise AI Projects Hit 70% Failure Rate as AMD, Red Hat, and SoundHound Rush Turnkey Platforms to Market →
• NVIDIA Robotics Platform Gains Traction in Global Manufacturing Through ABB, Universal Robots Deals →
• Open-Source AI Models Gain Ground Against Big Tech in Global Development Race →
• Open-Source AI Models Match Big Tech Performance as Global Sovereignty Concerns Reshape Industry →
• Pentagon Bans Claude AI as US, India, China Build Rival Tech Ecosystems →