Xuankun Rong

multimodal large language model (2) in-context learning (1) conversational ai (1) backdoor attack (1) vision-language model (1) large vision-language model (1) multi-turn dialogue (1) visual token (1) backdoor detection (1) multi-turn conversation (1) automated evaluator (1) semantic perturbation (1) inference-time defense (1) attention hijacking (1) test-time purification (1) confidence drift (1) benchmark evaluation (1) fine-tuning as a service (1)