AI安全 (2 articles)

Technical Concerns Behind Gemini 2.5 Pro's Dramatic Stability Decline

This week's evaluation data reveals Gemini 2.5 Pro's stability score plummeted from 54.0 to 31.2, a 42.2% drop, exposing serious issues in maintaining consistent output quality while other metrics improved.

In-Depth Analysis: From DeepSeek to Gemini, How to Build an Impregnable Defense Against "Model Distillation"?

This article analyzes the DeepSeek model distillation incident and proposes a comprehensive multi-layered defense system against distillation attacks, including API-level controls, output watermarking, and architectural protections.