一作ACL 2026Poster· Natural Language Processing
Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors
Rui Yin · Tianxu Han · Naen Xu · Changjiang Li · Ping He · Chunyi Zhou · Jun Wang · Zhihui Fu · Tianyu Du · Jinbao Li · Shouling Ji
摘要
Safety-aligned large language models (LLMs) are increasingly deployed in real-world pipelines, yet this deployment also enlarges the supply-chain attack surface: adversaries can distribute backdoored checkpoints that behave normally under standard evaluation but jailbreak when a hidden trigger is...