Tong Li

Network traffic alone can reveal LLM agent risks and granular behaviors, but current analysis methods fail to distinguish subtle malicious activities or semantically similar scenarios.

Parent-Conditioned Drafting can boost LLM inference speed by up to 29.5% while increasing the effective acceptance length of generated outputs.

Achieving state-of-the-art one-step image edits by cleverly balancing aggressive updates with conservative preservation techniques could redefine the capabilities of diffusion models.

Shopping agents struggle to achieve even 60% accuracy in uncovering hidden shopper intents across complex tasks, revealing critical gaps in current AI capabilities.

BBR-Copilot transforms BBR from a bulk data algorithm into a robust solution for live streaming, overcoming its critical bandwidth estimation flaws.