Tianwei Zhang

TransTroj enhances backdoor attacks, improving effectiveness and resistance to detection.

A new method aims to safeguard videos from harmful edits and manipulation.

A new method reveals backdoor attack threats in machine learning without sensitive data access.

A study reveals new techniques for backdoor attacks on language models with minimal impact.

A new method to improve safety tests for autonomous vehicles through generated scenarios.

A new system enhances the training of large language models with long sequences.

Examining the balance between geolocation technology and user privacy.

This study presents a new automated method for identifying spectral lines in space data.

A simple method to detect hidden cameras using affordable devices.

Disco4D generates realistic 3D human figures from a single image.

Real-time mapping helps robots navigate their environment safely and efficiently.

New methods help robots learn from past and future experiences effectively.

Discover how reinforcement learning refines large language models for better human interaction.

Learn how new watermarking methods safeguard your images online.

Engorgio exploits language models, raising concerns about service reliability and performance.